Analytical and Bioanalytical Chemistry https://doi.org/10.1007/s00216-021-03586-z REVIEW Metabolomics for personalized medicine: the input of analytical chemistry from biomarker discovery to point-of-care tests Florence Anne Castelli 1,2 & Giulio Rosati 3 & Christian Moguet 1 & Celia Fuentes 3 & Jose Marrugo-Ramírez 3 & Thibaud Lefebvre 1,4,5 & Hervé Volland 1 & Arben Merkoçi 3 & Stéphanie Simon 1 & François Fenaille 1,2 & Christophe Junot 1,2 Received: 22 June 2021 / Revised: 24 July 2021 / Accepted: 27 July 2021 # The Author(s) 2021 Abstract Metabolomics refers to the large-scale detection, quantification, and analysis of small molecules (metabolites) in biological media. Although metabolomics, alone or combined with other omics data, has already demonstrated its relevance for patient stratification in the frame of research projects and clinical studies, much remains to be done to move this approach to the clinical practice. This is especially true in the perspective of being applied to personalized/precision medicine, which aims at stratifying patients according to their risk of developing diseases, and tailoring medical treatments of patients according to individual characteristics in order to improve their efficacy and limit their toxicity. In this review article, we discuss the main challenges linked to analytical chemistry that need to be addressed to foster the implementation of metabolomics in the clinics and the use of the data produced by this approach in personalized medicine. First of all, there are already well-known issues related to untargeted metabolomics workflows at the levels of data production (lack of standardization), metabolite identification (small proportion of annotated features and identified metabolites), and data processing (from automatic detection of features to multi-omic data integration) that hamper the inter-operability and reusability of metabolomics data. Furthermore, the outputs of metabolomics workflows are complex molecular signatures of few tens of metabolites, often with small abundance variations, and obtained with expensive laboratory equipment. It is thus necessary to simplify these molecular signatures so that they can be produced and used in the field. This last point, which is still poorly addressed by the metabolomics community, may be crucial in a near future with the increased availability of molecular signatures of medical relevance and the increased societal demand for participatory medicine. Keywords Metabolomics . Biomarkers . Personalized medicine . Precision medicine . Point-of-care tests . Immunoassays . Biosensors Published in the topical collection Recent Trends in (Bio)Analytical Introduction Chemistry with guest editors Antje J. Baeumner and Günter Gauglitz. “Omics” analyses are based on the use of large-scale data * Christophe Junot christophe.junot@cea.fr production techniques (such as nucleic acid sequencing or mass spectrometry). These, coupled with data mining and in- 1 Université Paris-Saclay, CEA, INRAE, Département Médicaments et terpretation tools based on mathematics, statistical analyses, Technologies pour la Santé (MTS), Gif-sur-Yvette cedex 91191, and bioinformatics, make it possible to analyze and under- France stand a system or living organism as a whole, at the different 2 MetaboHUB, Gif-sur-Yvette, France levels of its biological organization (i.e., genomics, tran- 3 Institut Català de Nanociència i Nanotecnologia (ICN2), Edifici scriptomics, proteomics, and metabolomics for the sets of ICN2 Campus UAB, 08193 Bellaterra, Barcelona, Spain genes, transcripts, proteins, and metabolites, respectively). 4 Centre de Recherche sur l’Inflammation/CRI, Université de Paris, Metabolomics has emerged in the late 1990s, following the Inserm, Paris, France development of proteomics [1]. It deals with the detection, 5 CRMR Porphyrie, Hôpital Louis Mourier, AP-HP Nord - Université identification, and quantification of the small molecular de Paris, Colombes, France weight compounds present in a given biological medium. Castelli F.A. et al. The metabolome includes all the compounds of low molecular abundances are generally provided in chromatographic peak weight (< 1000 Da or < 1500 Da according to the definitions) areas. Those peak areas highly depend on the LC-HRMS con- that are present in a biological sample, and excludes biological ditions and thus remain linked to a dataset and are difficult to polymers such as proteins or nucleic acids [2]. Such metabo- be directly compared with those from other metabolomics lites can be molecules involved in the ubiquitous reactions by profiling experiments performed at different time periods or which the cells of an organism produce and use energy, as for within different laboratories. The most used detection example amino acids, sugars, nucleotides, or fatty acids. Some methods for global metabolomics approaches are NMR, GC/ other kinds of metabolites can be synthesized by any given MS, and liquid chromatography coupled to high-resolution biological species for achieving particular biological func- mass spectrometry (LC-HRMS). These untargeted ap- tions, as for example glucocorticoids or neurotransmitters in proaches are used especially in the first steps of biomarker mammals, or alkaloids in plants. Xenobiotics (i.e., chemicals discovery. Conversely, targeted approaches are focused on a which are found in living organisms, but not produced by limited number of compounds, belonging to a given chemical them) such as drugs and pollutants and their metabolites are class or metabolic pathway. They can be more sensitive than also part of the metabolome since they are present in biolog- untargeted approaches and can provide, if needed, absolute ical media and can be detected by the analytical chemistry quantification with results expressed in molarity units. They tools that are used for metabolomics [3]. This highlights that can be used for mechanistic purposes (in the frame of metabolites are of different origins and come not only from the fluxomic experiments, for example) or for biomarker valida- cellular metabolism, but also from the microbiota, food and tion following untargeted approaches [6, 7]. drinks, drug intake, and environment. This makes metabolo- Metabolomics workflows, including sample preparation, mics an efficient tool to track interactions between a living MS and/or NMR analyses, data pre-processing, statistical organism and its environment. However, many metabolites analyses, and data visualization, have been developed since are still uncharacterized due to their large chemical diversity the 2000s and have now reached a certain level of maturity. and to insufficient knowledge about metabolism. As a matter Nowadays, metabolomics is considered as a tool in its own of fact, it is estimated that less than 5% of the features detected right and is used in systems biology projects in many biolog- in biological media using mass spectrometry-based metabolo- ical fields, such as environmental research [8], plant science mics methods are annotated [4]. [9, 10], nutrition, animal and human health [11, 12], etc. In Metabolomics is thus far more complex than the metabolic particular, this is the case in the field of biomedical research, pathways displayed in metabolic biochemistry textbooks. It is for which systems biology has given rise to systems medicine, indeed impossible to know exactly how many metabolites a systems approach to health and diseases paving the way to compose the metabolomes of living organisms, whereas the- personalized medicine. oretical proteomes can be more readily inferred from genomic Personalized medicine highlights the importance of the in- data. Furthermore, as metabolites exhibit a huge chemical di- dividuals’ characteristics in the response to treatment. This versity, ranging from polar and hydrophilic compounds such concept has been developed with the improvement of our as sugar derivatives to apolar and hydrophobic molecules such knowledge, which makes it possible to define pathologies as lipids, there is no universal method for metabolome analy- more precisely. According to Leroy Hood, one of the pioneers sis. Consequently, the detection of metabolites relies on com- of this approach, personalized medicine promises to (i) pro- plementary methods that have to be run in parallel to achieve vide deep insights into disease mechanisms, (ii) make blood a optimal metabolome coverage [5]. The two main technologies diagnostic window for viewing health and disease of an indi- for metabolomics data production are based on nuclear mag- vidual, (iii) stratify complex diseases into subtypes, (iv) pro- netic resonance (NMR) spectroscopy and mass spectrometry vide new approaches to drug target discovery, and (v) gener- (MS). The latter can be used as a standalone technique (i.e., ate metrics for assessing wellness. Thus, medicine aims at direct introduction mass spectrometry) or coupled with gas being preventive, predictive, personalized, and participatory, chromatography (GC-MS), liquid chromatography (LC- referred to as the concept of “P4 medicine” [13]. This ap- MS), or capillary electrophoresis (CE-MS). proach is particularly relevant in the field of chronic and There are two main kinds of metabolomics approaches: non-communicable diseases for which it is often difficult to targeted and untargeted (or global) ones. Untargeted ap- grasp the transition from the healthy to the disease status. In proaches seek to detect as many compounds as possible in this context, Sagner et al. have proposed a “P4 health contin- samples. Such compounds can be known metabolites, puta- uum model” with 4 stages of health, namely the A, B, C, and tively annotated metabolites, or unknown ones. In this case, D stages, corresponding respectively to an apparently healthy metabolite concentrations are given in a semi-quantitative state, the emergence of chronic disease signs (such as elevated manner (relative quantification). This means that metabolite blood pressure or dyslipidemia), the emergence of chronic concentrations are expressed as arbitrary units or ion intensi- disease symptoms, and the confirmed chronic disease diagno- ties. In untargeted MS-based metabolomics, metabolite sis, respectively [14]. Metabolomics for personalized medicine: the input of analytical chemistry from biomarker discovery to... Research and development in biomarker discovery are Metabolomics for personalized medicine: central in modern healthcare for personalized and preci- more standardized and sharable sion medicine. Omics approaches can be regarded as metabolomics datasets are still required particularly relevant and useful tools to identify new mo- lecular biomarkers or sets of biomarkers to improve the Several recent reviews have already addressed clinical appli- diagnosis and prognosis of various diseases as well as to cations of metabolomics in various fields such as oncology, evaluate treatment efficacy. In this context, metabolomics cardiology, neurology, diabetes, kidney and liver diseases, represents an attractive strategy for profiling a large pan- and also response to treatments (aspirin, simvastatin, or anti- el of low molecular weight molecules in patient samples hypertensive drugs) [16, 18–21]. and for pointing out relevant molecules closely related to In their article entitled “Metabolomics for the masses: the (patho)physiological conditions and treatment response future of metabolomics in a personalized world,” Trivedi et al. phenotypes. However, over the last ten years, the number have discussed on the current positioning of metabolomics of biomarkers derived from omics-based approaches, ap- and on its future in a context of personalized medicine and, proved by regulatory agencies and used in clinical set- more generally speaking, healthcare. They have listed around tings, remains far from expectations [15]. This often one hundred metabolomics studies proposing biomarkers of makes people perceive metabolomics and other omics various pathologies and published between 2000 and 2017. as over-promising and/or under-delivering approaches The authors pointed out limitations at different levels: data when applied to clinical questions [16]. Despite the pub- production with a lack of interoperability and methodological lication in the last two decades of more than 2000 sci- validation, poor experimental design with many studies deal- entific papers using MS-based metabolomics for human ing with less than one hundred subjects (leading to a lack of disease diagnostics, no diagnostic test based on metabo- statistical robustness and validity), with the absence of lomics has yet reached the clinics [17]. There are several replication/validation cohorts. Finally, stating that mass reasons and explanations for this observation, the most spectrometry–based approaches are not suitable for large- frequently cited being the difficulty to integrate scale screening of a very large population due to their low multiscale biological information to generate knowledge throughput and high price, they consider alternative analytical and predictive models [13, 16], the inappropriate design chemistry tools accessible to non-specialist end users, such as of clinical trials with an often too small number of pa- lateral flow devices, dipstick approaches, or breath measure- tients, together with a lack of validation cohorts [16, 18], ments of volatiles [18]. but also due to issues at the level of data acquisition, i.e., The same conclusions are presented in the position paper of analytical chemistry. Indeed, the lack of standardization Pinu et al., which summarizes the discussions on translational of the data production methods, together with the expres- metabolomics undertaken during the peer session of the sion of results in a semi-quantitative manner, are fre- Australian and New Zealand Metabolomics conference in quently highlighted as factors preventing the sharing 2018. In addition to issues related to the lack of standardiza- and reuse of metabolomics data, and their integration tion of data acquisition protocols and to partial identification into multi-omics models [16, 18]. Anyway, while a and too limited annotation of datasets, these authors also em- growing number of studies report on metabolic signa- phasized the insufficient recognition of metabolomics by tures for the diagnosis and monitoring of pathologies, funding agencies, the difficulty of gathering multiple fields responses or non-responses to treatments, it is now nec- of expertise within a given group, and the necessity to develop essary to consider converting these complex and routine tests and portable devices [16]. multiparametric signatures into reliable assays, with ap- A search in the PubMed database (March 2021) over the propriate costs of sample analysis, that could be run in 2015–2020 period with keywords related to the main omics medical biology laboratories and even point-of-care tests, techniques, combined with the terms “disease*,” “*marker* which are part of the participative dimension of person- or signature*,” and “patients,” and restricted to the occurrence alized medicine. These aspects, which fully deal with of these words in titles and/or abstracts, yielded 4874 publi- analytical chemistry, are still poorly addressed within cations. As shown in Fig. 1(A), the bulk of these studies in- the metabolomics research community. volves transcriptomic analyses. Of these, the contribution of In this context, this review deals with the main challenges metabolomics or lipidomics analyses amounts to 13%, that of linked to analytical chemistry that need to be overcome to proteomic analyses to 15%, while approaches linked to foster the implementation of metabolomics in personalized microbiome and multi-omic analyses are emerging. medicine and in clinical practice, from data production for Figure 1(B) displays the number of publications mentioning biomarker discovery and validation, to the translation of meta- the use of metabolomics or lipidomics alone or with other bolomics signatures into assays for medical laboratories and omics approaches. A fourfold increase in the number of pub- point-of-care tests. lications referring to metabolomics has been observed Castelli F.A. et al. Fig. 1 Omics technologies for biomarker discovery in the medical field. (A) Pie chart displaying the relative contribu- tions of the various omics ap- proaches for the discovery of biomarkers of diseases over the 2015–2020 period. The PubMed database was inquired (March 2021) with the following keywords occurring in titles and/ or abstracts: “disease*,” *marker* or signature*, patients; and “metabolom* or lipidom*,” “transcriptom* or gene expres- sion,” “proteom*,” “microbiom*” and “multi-omics,” excluding re- view articles. (B) Number of 1% publications related to metabolo- mics and/or lipidomics alone or 6% 13% combined with other omics from 2015 to 2020 15% 65% metabolom* OR lipidom* proteom* transcriptom* OR gene expression microbiom* multi-omics between 2015 and 2020. Although metabolomics analyses are biological samples. A large number of projects are related to still mainly used alone, there is a trend toward increased use of diseases such as cancer (182 projects) or diabetes (63 multi-omics approaches in 2020 (62 publications out of 187). projects). It should also be noted that spectral data Finally, among the 643 publications related to metabolomics warehouses such as MassIVE (https://massive.ucsd.edu), a or lipidomics, only 15 of them mention the use of a validation community resource developed by the NIH center for cohort to confirm the molecular signatures obtained. computational mass spectrometry for the free exchange of At the same time, data repositories have been developed, mass spectrometry data, are likely to host metabolomics making publicly available projects and studies including raw datasets (404 out of 11,065, March 2021). data together with analytical and biological metadata neces- The development of data warehouses has been facilitated sary for statistical analysis, as well as processed and annotated by the efforts made within the community to standardize the datasets. The two main data warehouses in the field of meta- different steps of the metabolomics analysis workflow. bolomics are MetaboLights [22, 23] and Metabolomics Initially, the Metabolomics Standards Initiative, set up by Workbench [24]. They allow sharing of protocols, assays, the Metabolomics Society in 2005, has addressed, through and even tools to perform analyses and meta-analyses on the working groups, various aspects related to ontology problems, datasets (for Metabolomics Workbench). At the time of writ- data exchange, biological metadata, chemical analyses (from ing this review, MetaboLights (https://www.ebi.ac.uk/ the reporting of analytical chemistry metadata to metabolite metabolights, accessed March 2021) contained 554 studies, identification status), or data processing and statistical analy- of which 30% deal with human studies (i.e., related to ses [25]. Regarding raw data sharing, vendor independent “homo sapiens” in the browse study menu) and about 18% standards have been proposed since the 1990s, and nowadays, are performed on human blood and urine samples (i.e., related XML (i.e., “eXtensible Markup Language”)–based formats to “blood,” “serum,” “plasma,” “whole blood,” and “urine” in such as mzML [26] and nmrML [27] are supported by vendors the browse study menu). Metabolomics Workbench (https:// and commercial and open-source software [28]. Concerning www.metabolomicsworkbench.org, accessed March 2021) metadata associated to studies, the ISA-Tab format (“ISA” for included 1370 studies of which 41% dealt with human Investigation, Study, Assay, and “Tab” for tabular) has been Metabolomics for personalized medicine: the input of analytical chemistry from biomarker discovery to... developed to collect and share both analytical chemistry and The study of Pang et al. deals with a comprehensive meta- biological metadata [29]. Finally, certified reference materials analysis of COVID-19 global metabolomics datasets. It was are provided by metrological institutions, which facilitate the achieved by using seven liquid chromatography coupled to implementation of international inter-laboratory studies [30]. mass spectrometry datasets obtained from six studies that This is, for example, the case for human plasma with the SRM were realized in three distinct countries. Five datasets were 1950, which is produced by the U.S. National Institute of obtained as raw data from public repositories or from the Standards and Technology (NIST) [31], and which has been authors of the studies, whereas the remaining two datasets used in several inter-laboratory tests [32–35]. Beyond the case consisted of annotated peak tables obtained from supplemen- of human blood samples, NIST urine reference materials have tary materials of publications. The authors implemented a been used in an inter-laboratory study involving NMR, GC- computational workflow to process the raw data coming from MS, and LC-MS untargeted metabolomics analyses [36], and different experiments and performed pathway enrichment and the development of a human stool reference material for meta- visual data mining, leading to metabolic signatures character- bolomics and metagenomics gut microbiome analysis is en- istic of the disease progression and clinical outcomes [40]. visaged, as highlighted in a workshop report [37]. In summary, progresses have been achieved in terms of However, despite all these advances, the use and reuse of technological advances for the production and analysis of public data from global metabolomics analyses in the frame of metabolomics data, and for the standardization and sharing meta-analyses remain uncommon. Table 1 shows the main of these data. These advances are reflected in the increasing meta-analyses published to date. They are all based on mass involvement of metabolomics data in clinical studies dealing spectrometry data, and some of them use software tools such with disease biomarker discovery, and in systems medicine as MetaboAnalystR 3.0 [38] or PAIRUP-MS [39], which en- research projects based on multi-omics approaches. able joint analyses of different projects or batches from raw However, global and non-targeted metabolomics analyses data and pathway enrichments. are still very little used, or even not at all, in routine care Table 1 Meta-analyses involving untargeted metabolomics-based Publication title Data Technology Software Reference approaches Comprehensive 7 datasets from 3 LC/HRMS MetaboAnalystR Pang et al., meta-analysis of countries, including 5 3.0 Metabolit- COVID-19 global raw datasets from es, 2021 metabolomics MetaboLights, [40] datasets MassIVE, and authors, and 2 annotated peak tables from 2 publications. 438 blood samples from 337 subjects Benford’s law and Datasets from 3 studies LC/HRMS No D'alessandro, metabolomics: a tale performed by the Transfus of numbers and blood author, no raw data Apher Sci, available, peaktable 2020 [194] available for one study Integrating untargeted 3 LC/MS datasets, no LC/HRMS PAIRUP-MS Hsu et al., Int metabolomics, raw data available, J Obes genetically informed one peaktable avail- (Lond), causal inference, and able (related to the 2020 [195] pathway enrichment software publication) to define the obesity metabolome MicroRNAs regulating Meta-analysis including LC/HRMS, No Wang et al., human and mouse microRNA-seq, LC/QQQ-- Int J Mol naïve pluripotency RNA-seq, and MS, Sci, 2019 metabolomics GC/MS [196] datasets; the metabolomics datasets are from a single published study; peaktables available; no raw data available Castelli F.A. et al. practices, especially in medical laboratories. It is therefore data scientists, each one knowing the constraints and require- necessary to pursue research activities in the field of data ments of their respective field. A successful metabolomics production and interoperability, as detailed below with a focus study requires some key elements, including but not limited on mass spectrometry–based metabolomics approaches. to (i) consistency of samples to be analyzed and compared (e.g., individuals matched for sex, age, weight, ethnicity, life- style factors, etc.; site and type of sample, sample shipment, Toward inter-operable and reusable storage, and handling), (ii) proper study design, (iii) proper metabolomics data for biomarker discovery: control groups and conditions, and (iv) sufficient sample size from appropriate project design and sample of compared groups to encompass inter-individual variability collection to confident identification and provide statistical power [44] (Fig. 2). and measurement of biomarker candidates Preanalytical considerations, study design The levels of me- The pipeline of biomarker development includes several key tabolites constituting the metabolome of any individual are stages, consisting of discovery, validation, and clinical trans- quantitatively affected by many factors such as disease, drug lation [41]. Each of these steps has its own limitations and can use, environmental exposures, or nutrition [6, 45]. Therefore, be improved. For instance, the discovery phase might suffer collection of data related to demographic, lifestyle, and phys- from a lack of standardized and validated methods, yielding iological factors (e.g., diet, gender, ethnicity, age, and BMI) is poor experimental reproducibility between laboratories. This necessary to limit inter- and intra-individual variations and is especially the case with MS-based metabolomics, whereas, identify potential confounding factors. This practice also en- although detecting less metabolites than LC-HRMS, NMR sures that appropriate data can be incorporated into the exper- can be more directly quantitative and can thus deliver more imental design and data analysis processes [46]. robust data than untargeted LC-HRMS-based approaches. Obviously, sample type/matrix must be consistent between Another reason is that clinical studies are not always appro- the patient samples and the control group (e.g., plasma priately designed for biomarker discovery, with biomarker EDTA). Although overall concordance in results and similar candidates not validated in independent cohorts and also not discriminatory ability can be observed when running plasma sufficient clinical phenotyping available [42, 43]. In this con- and serum metabolomics [47], serum appeared to show higher text, in the following paragraphs, we will review and discuss levels of some metabolites as a potential consequence of the the recent findings in LC/MS-based metabolomics that can fill clotting process, as it is exemplified in the study of Wedge these gaps by pinpointing the need of the following: et al., which deals with the metabolomics analysis of 29 small- cell lung cancer patients [48]. Some metabolites can also read- – Large multi-center cohorts as well as validation cohorts to ily degrade or interconvert during sample preparation or dur- increase statistical power and biomarker specificity and ing temporary storage at room temperature before analysis avoid confounding factors [49–51]. Kamlage et al. reported on the impact of blood and – Improved metabolome coverage and metabolite identifi- plasma processing on the concentration of metabolites. They cation confidence level (e.g., thanks to optimized and observed significant metabolite concentration variations (in validated acquisition workflows, dedicated spectral the 4–19% range) in the case of hemolysis or short-term stor- databases) age of plasma samples at room temperature or cooled on wet – Standardized data production workflow with improved ice, and also some minor consequences when samples were robustness (e.g., with QCs, batch-to-batch consistency/ contaminated with buffy layer or in case of micro-clotting normalization, large-scale quantification) and capability [52]. Caution has to be paid to avoid these issues. of automated interpretation of the huge amount of data Different study designs (e.g., cross-sectional, cohort, case- generated control studies) have their own strengths and weaknesses, and – Linked untargeted and targeted quantitative approaches their applicability depends on the clinical question of interest for proper analytical validation of biomarker candidates [53]. Moreover, the sample size of groups should be similar and large enough in a comparison to provide sufficient statis- tical power. However, it is important to note that some cohort The issue of design of experiment regarding cohort studies may use more controls than cases in their experimental samples: toward more standardization design, which might imply the development and use of spe- cific statistical tools for the treatment of the resulting metabo- Metabolomics is a question-driven method. Thus, the prereq- lomics data. In the case of expected high inter-individual var- uisite of a successful metabolomics experiment is a well- iability, a larger sample cohort would even be required. The defined clinical question, which would unarguably imply ac- issue of statistical power related to sample size for the appro- tive discussions between physicians, analytical chemists, and priate design of metabolomics studies has already been Metabolomics for personalized medicine: the input of analytical chemistry from biomarker discovery to... Fig. 2 The road to successful biomarker discovery addressed through modeling approaches using pilot studies Indeed, a common issue in metabolomics-based biomarker [54], or even when pilot studies are not available [55]. discovery is the absence of replication and insufficient sam- However, there is still a lack of reference procedures to ad- pling, with many studies involving a single cohort of limited dress this point [56]. Thus, experts in data production and size, i.e., often fewer than 100 samples [60]. Replication of the statistical analysis should be involved in experimental design, results on biomarker performance using independent multi- together with the sponsors of the study [57–59]. They should center cohorts is needed to ensure proper biomarker validation pool their expertise to ensure the most standardized and robust for further transfer to the clinics [18]. Collaborative networks experimental design. Important points to be discussed and and access to well-characterized bio-banked samples are often agreed are the following: (i) a clear definition of the research needed to perform such large-scale studies. question, an as-precise-as-possible description of clinical characteristics of each group to be compared (in order to avoid Standardized sample preparation As the level of any unwanted inter-group variability); (ii) a “true control” comparability/variability is the critical point to be controlled group (although it is often complicated to collect samples for standardizing metabolite concentration in the frame of from healthy people in hospitals); (iii) an estimation of the untargeted metabolomics experiments, standard operating minimal sample size that is required to address the biological procedures (SOP) for sample collection and metabolite extrac- question, the number of sample aliquots needed for the study; tion are essential [57, 61, 62]. The review of Kirwan et al. [57] and (iv) the creation and sharing of a single exhaustive file of gives a valuable overview about well-defined and validated sample metadata with all known technical and clinical vari- protocols for the collection of samples for metabolomics ables for each sample (sample name, tube labeling or barcode, research. sample box, location in the box, collection origin, collection Kirwan et al. listed several already published protocols for date, gender, age, clinical treatment, diet, disease, BMI, used collection and storage of numerous biofluids and tissues. anticoagulant, ethnicity...). Such a file could help to better Important insights are synthesized in an informative table define the experimental groups in terms of homogeneity, highlighting the crucial points of the protocols related to meta- while avoiding any confusion and limiting the occurrence of bolomics: (i) collection methods, (ii) range of temperature possible confounding effects (Fig. 2). before processing, (iii) storage conditions and reported conse- quences if not followed, and (iv) observed confounding ef- Discovery and (pre)validation cohorts As mentioned above, fects [57]. untargeted LC-HRMS metabolomics methods are mainly Before implementing an extraction protocol chosen as the used for biomarker discovery. The resulting biomarker candi- most appropriate to address the biological question of interest, date data should be considered preliminary until their valida- a crucial step in the standardization process consists in sample tion on a larger cohort with the use of a similar untargeted normalization in terms of metabolite concentration before data workflow or a targeted quantitative metabolomics approach. acquisition [62]. Regarding serum and plasma analyses, a Castelli F.A. et al. consensus normalization way is to analyze the sample using a achieve broader metabolite coverage, samples are often ana- fixed sample volume [61, 63]. As for a biomedical analysis, lyzed several times by complementary LC-HRMS(/MS) the sample must be collected in a fasting state to minimize methods [63, 75–77]. A large portion of the detected signals unwanted sources of variability on the metabolome. For hu- remains, however, structurally uncharacterized, and thus, me- man fecal samples, Karu et al. also described several proposed tabolite identification still represents a major bottleneck in protocols for human fecal metabolomics [64]. We advise to metabolomics [4, 78]. Detecting as many metabolites or me- normalize on the dry weight of the freeze-dried stool sample to tabolite features as possible with the objective of obtaining avoid variability induced by differences in water content from maximal biochemical information is a general tendency. one fresh stool sample to the other [65]. Regarding urine However, confidently identifying and measuring true and metabolomics studies, the need of normalization due to diure- unique metabolites is a completely different objective that is sis variations is well known and it has been already addressed absolutely required to obtain both reliable biomarker candi- in several publications. Ideally, all urine samples in a study dates and meaningful biological information readily sharable should be collected over a period of time such as 24 h, know- between laboratories. Therefore, broad metabolome coverage ing that the metabolic content of the sample can be impacted makes sense only if metabolites are annotated/identified at a by bacterial growth and chemical stability issues [50]. The high confidence level. Of course, this might sound frustrating most used normalization protocols rely on creatinine, osmo- or can be erroneously linked to poor methodological perfor- lality, total useful signal (TUS, post-acquisition normaliza- mances since this often yields limited sets of metabolites (e.g., tion), and specific gravity measurements [66–71]. Although ~ 200), instead of few thousands of metabolite features or no clear recommendation has emerged, the limitation of using elemental compositions. Such “limited” number of metabo- creatinine, encountered in many pathological contexts, has lites can already provide key insights into relevant clinical been highlighted many times, while osmolality and specific questions. For example, the robust monitoring of 137 metab- gravity appear to be the most reliable normalization protocols olites in the blood of 800+ patients with acute decompensation [68, 71, 72]. of cirrhosis with/without acute-on-chronic liver failure For tissues, it is often difficult to weigh accurately a small (ACLF) provided unprecedented insights into the biochemical piece of frozen sample. We advise a standardization related to mechanisms, underlying the development of the ACLF syn- the measurement of the total protein concentration. The pro- drome and also a 38-metabolite blood fingerprint specific for tein concentration is thus measured in the pellet obtained after ACLF that revealed mitochondrial dysfunction in peripheral the protein precipitation step during the metabolite extraction organs [79, 80]. (e.g., by using bicinchoninic acid “BCA” protein assay). While for bacteria the normalization can be made based on Confident annotation and identification of metabolites As the optical density measurement [73, 74], normalization of mentioned above, accurate identification and monitoring of eukaryote cell samples before acquisition can be done on cell metabolites are prerequisites to achieve measurements’ repro- numeration (which may lack of accuracy), or more confident- ducibility across laboratories and among countries. Confident ly by using total protein concentration or DNA concentration metabolite identification in complex biological matrices re- in the extract. To conclude, it is essential to normalize the quires the combination of several information lines exploited concentration of metabolites present in the final extract and in conjunction. High mass resolution (> 100,000, M/ΔM, full before acquisition to detect only the metabolic variations re- width at half maximum) and high mass measurement accuracy lated to the clinical parameter being monitored. Post- (< 1 ppm) allow for the measurement of isotope pattern and acquisition normalization can also be implemented, for exam- isotope fine structure, including the distinction of isobaric iso- ple, by using the TUS (as mentioned above for normalization topes [81–83]. Complementarily, accurate retention time(s) of urine samples) [68]. and MS/MS spectra can provide high confidence in metabolite identification when matched to reference data included in The issue of metabolome coverage and accurate chemical/spectral libraries. Among those complementary in- measurement formation lines, one of the most valuable elements to confirm metabolite annotation or reduce the list of possible annotations Obtaining an exhaustive picture of the metabolome is highly is the acquisition of fragmentation spectra and their compari- desirable to increase the likelihood of getting the best bio- son to reference MS/MS spectra included in reference mass marker or set of biomarkers. Although major developments spectral libraries [82]. Differently from proteins, fragmenta- have already taken place, we are still a long way from getting a tion of metabolites under low-energy conditions is relatively comprehensive coverage of all the metabolites. In the absence unpredictable (at least with a high confidence level). of a universally accepted procedure for biomarker discovery Therefore, the most relevant mass spectral databases for de- by mass spectrometry–based metabolomics, each data pro- finitive identification of metabolite biomarker candidates are duction facility uses its own optimized procedure [18]. To those obtained from pure authentic standards, and are thus Metabolomics for personalized medicine: the input of analytical chemistry from biomarker discovery to... unfortunately limited by their (most often commercial) extended to unknown signals returned by meta-analysis at availability. the pathway level with software tools such as Mummichog [97]. Databases are essential but often reflect only what is identi- fied and commercially available Most widely used (public and Reporting high-quality and robust data in the frame proprietary) spectral databases include MassBank [84], of untargeted MS-based metabolomics studies HMDB [85], GNPS [86], MoNA (https://mona.fiehnlab. ucdavis.edu/), LIPID MAPS [87], NIST 20 (https:// General quality assurance and quality control practices chemdata.nist.gov/dokuwiki/doku.php?id = chemdata:msms) Quality assurance (QA) and quality control (QC) are indis- , METLIN [88], and mzCloud (https://www.mzcloud.org/). pensable processes in research and critical in metabolomics For a more detailed comparison and discussion of these to obtain high-quality and FAIR data [58, 59, 98, 99]. databases, we recommend these articles in the field [89, 90]. QA deals with processes planned and performed before However, in brief, procedures for data collection and curation, sample collection to fulfill predefined quality requirement. instruments used (e.g., Orbitrap, Q-TOF), fragmentation The main processes for metabolomics experiments are staff conditions (e.g., resonant and/or non-resonant conditions, training, preventive instrument maintenance, analytical number of collision energies, MSn), and type of molecules methods validation, and calibration of analytical tools and differ from one database to the other. Presently available spec- verification of their performance. tral databases are therefore not strongly overlapping [90], Quality control deals with techniques and activities used to which underlines their complementarity as exploited by recent measure and report if these QA requirements have been met studies for larger metabolite identification [91]. Interestingly, during and after data acquisition. It can include (i) run order a European proposal for quality control and quality assurance randomization to control unwanted variation and unwanted of tandem mass spectral libraries has been recently published correlation; (ii) internal standards added to samples in order and also reported that Q-TOF and Orbitrap-based instruments to report the quality of data within and between batches; (iii) yielded comparable MS/MS spectra [92]. As a representative blank samples to assess the occurrence of contaminations, example, the MS/MS spectrum of taurocholic acid as acquired interferences, artifact feature, carry-over phenomenon, and in the positive ion mode on a first-generation Thermo Q- matrix effect (if internal standards are added); and (iv) QC Exactive instrument (at a normalized collision energy samples (including diluted QC in order to highlight artifactual “NCE” of 20%) has been successfully matched to spectra features). stored in the MoNA database and previously recorded both QC samples should be representative of the qualitative and on a Q-Exactive HF and a Waters Q-TOF II instruments used quantitative composition of the samples to be analyzed. The under different fragmentation conditions (Fig. 3). This opens samples can be obtained by pooling aliquots of each original the door to spectral databases generated in the FAIR (i.e., sample and then following the same sample preparation pro- “Findable, Accessible, Interoperable, Reusable”) data context tocol of the biological samples. Proposed in 2006 for the first [93] also with a standardized way of describing the observed time for untargeted metabolomics studies in order to check for ions [94]. data quality [100], QC samples have been and are used in Although covering more and more metabolites, not all me- numerous studies and several publications have highlighted tabolites found naturally in biological matrices are currently their importance in the metabolomics workflow [59, 61, 98, commercially available to feed those databases, which clearly 99, 101]. Broadhurst et al. have reported on the different kinds limits their expansion. As a matter of fact, metabolomics re- of QC samples, and their respective advantages and limita- searchers usually focus on metabolites that are present in da- tions [98]. QC samples are usually injected several times at tabases or websites. Although very exciting for the analytical the beginning of each batch analysis in order to equilibrate the chemist, the exploration of unknown metabolites, the so- analytical platform, and then every 5 to 10 biological samples called dark matter [95], is one of the biggest challenges and in order to control the stability and performance of the analyt- a research field on its own and is therefore still insufficiently ical platform, and its reliability during the course of analysis. explored. Therefore, acquisition and thorough annotation (or They have also a crucial role in data pre-processing, as em- putative identification) of MS/MS data from biologically rel- phasized below. evant and recurrently observed metabolite features absent from databases and chemical provider catalogs is required The need for reproducible data (pre)processing workflows for their further implementation and storage in specific shared and precisely documented metabolite identification proce- databases. Such a strategy has been recently used to build dures This includes data preprocessing and processing, the spectral libraries of unidentified but annotated recurrent spec- reporting of metabolite identification, and publication and da- tra derived from NIST urine samples [96], and could be ta sharing. Castelli F.A. et al. Fig. 3 Head-to-tail comparison of evaluated versus reference MS/ MS spectra of taurocholic acid obtained under non-resonant con- ditions. Evaluated MS/MS spec- tra (blue color) were obtained in the positive ion mode on a Thermo Q-Exactive instrument (NCE 20%). Reference spectra (red color) are stored in the MoNA database and were re- corded in the positive ion mode on a Q-Exactive HF (NCE 20– 30–40%) and a Waters Q-TOF II instrument (20 eV). [M+H]+ ion at m/z 516.2974. Spectral matching was performed using the MS-DIAL version 4.12) soft- ware [208]. DP, dot product; rev. DP, reverse dot product The data-preprocessing step deals with the implementation metabolites features contained in peak tables into biologically of a peak-picking (i.e., detection and integration of LC/MS interpretable data. The way of reporting metabolite identifica- peaks, i.e., features), feature alignment and feature integration tion has already been highlighted as a key issue. The Chemical workflow. This is a key step in untargeted metabolomics, Analysis Working Group of the Metabolomics Standards which leads to data matrices that have to be cleaned and an- Initiative has thus proposed four levels of confidence for me- notated before being subjected to statistical analyses. Software tabolite identification [105]: packages dedicated to NMR, GC-MS, or LC-HRMS are avail- able for that, and their description is beyond the scope of this – Level 1: fully identified compounds based on at least 2 review. There is a need to standardize these pre-processing orthogonal (i.e., independent) data related to an authentic steps [102], but most importantly to standardize the sharing standard analyzed in the same experimental conditions. of processed data so that they can be re-analyzed by other – Level 2: putatively annotated compounds based on char- investigators to advance in the development of these tools. acteristic physicochemical properties or spectral similari- This can be achieved by integrated software tools that com- ty with spectral libraries. In this case, there is one candi- bine several data treatment steps, from data preprocessing to date, but the authentic standard is not available for statistical analyses, such as workflow4metabolomics [103] or confirmation. Metaboanalyst [104], the latter also proposing some data vi- – Level 3: putative characterized compound classes, based sualization and metabolic pathway enrichment tools. Such on characteristic physicochemical properties of a chemi- metabolomics data-processing platforms enable to store and cal class (i.e., acylcarnitine derivatives, or share all the parameter values related to data pre-processing sulfoconjugates). In this case, several candidates are pos- and processing. sible and it is not possible to highlight a precise chemical Dataset annotation and metabolite identification are crucial structure. steps enabling to convert hundreds to thousands of – Level 4: unknown compounds. Metabolomics for personalized medicine: the input of analytical chemistry from biomarker discovery to... Although they are simple to implement, these identification some cases, the strategy of the “pooled QC sample” might not levels have limitations since they cannot avoid ambiguities. be the most relevant. For instance, when analyzing some par- For example, the use of at least two orthogonal data together ticularly large cohorts, the amount of QC sample available can with that of a standard, as it is required for a level 1 metabolite become limiting; while for projects involving rare diseases or identification, does not always lead to an unambiguous iden- longitudinal studies, the exact number of batches to be re- tification, as observed with optical isomers such as enantio- ceived can be unknown or the samples are not all available mers, for example [106]. Although improvements have been at the beginning of sample preparation and analysis. In such proposed with the addition of a fifth level [107], or with the situations, a commercially available plasma LTR QC sample proposal of metabolite identification metrics with a quantita- from NIST (see below) can represent a reliable alternative. tive scoring system based on the sum of all the data types There are several key issues associated with these LTR QC supporting the identification [108], there is a need to propose samples. The first one deals with how representative such a metabolite identification reporting system that focuses fur- samples are, in terms of metabolites and concentration ranges. ther on chemical structures. As an example, the standard reference material (SRM) dedi- cated to metabolites in human plasma (SRM 1950) is prepared Application to different types of medical cohorts from 100 donors, with an equal number of men and women of 40 to 50 years of age, selected to be representative of the Depending on the level of complexity of the study, different ethnic distribution of the US population [31]. However, it is kinds of quality control procedures should be applied to the not established whether or not this SRM will be appropriate metabolomics data production workflow in order to achieve a for studies performed on other populations, or on patients sufficient degree of standardization. Thus, non-targeted meta- affected by overt diseases or metabolic disorders. Further bolomics studies performed with MS technologies can be studies are needed to address these issues. classified into three categories: The second one is their long-term stability. Indeed, stability - (a) Studies dealing with a sample set analyzed in a single studies are challenging to implement in the case of non- batch, in the same laboratory, and with the same instrument targeted and semi-quantitative approaches, especially those - (b) Studies dealing with a large cohort requiring to be based on the use of mass spectrometry, due to the impossibil- analyzed in several batches, over a long period of time, in ity to ensure the stability of the metabolites contained in the the same laboratory QC samples that are used for inter-batch normalization. - (c) Studies dealing with a large cohort and requiring sev- However, a few studies dealing with untargeted metabolomics eral laboratories approaches have been published on this issue. For example, All the QA and QC procedures previously detailed below Laparre et al. have evaluated the impact of the storage tem- apply in terms of experimental design, sample preparation, perature (+ 4 °C, − 20 °C, − 80 °C, and freeze-dried stored at − data production, and pre-processing. The most important ele- 80 °C) and the storage duration (5 to 144 days) on the bovine ment is to use QC samples and to analyze them at regular urinary metabolome by using liquid chromatography coupled intervals in order to guarantee signal stability over the duration to high-resolution mass spectrometry [114]. The authors fo- of the experiment. cused on 200 identified metabolites contained in their spectral database. Normalization was performed by dividing each fea- Studies dealing with a sample set analyzed in a single batch, ture’s intensity recorded on the different days by the corre- in the same laboratory and with the same instrument This sponding feature’s intensity recorded in the freshly collected case deals with cohorts ranging from 200 to 500 samples, urine sample at day 0 for every subject. Furthermore, scaling depending on the kind of sample preparation protocols and factors were calculated by dividing the average intensities instruments used. It is the simplest one because recommenda- recorded in − 80 °C stored samples per feature and for each tions on how to standardize the different steps of the batch by the corresponding intensities recorded in the refer- untargeted metabolomics workflow and quality management ence batch. By these means, they found that urine metabolic practices (from the experimental design of the study to the profiles are altered starting from 5 days when stored at + 4 °C, submission of data into public repositories) are available and and after one month at − 20 °C. The temperature of − 80 °C have been the subject of many publications [30, 57–59, 61, 62, was considered as the most convenient urine long-term stor- 98, 99, 105, 109–113] (Table 2). age condition. In another study, Palmer et al. have investigated the 12-month stability of dried blood spots (DBS) and dried Studies dealing with large cohorts requiring to be analyzed in urine spots (DUS) at different storage temperatures (− 20, + 4, several batches, over a long period of time, in the same lab- and + 21 °C) and compared it to plasma and urine biofluids oratory In this case, in addition to the already described QA stored at the same storage temperatures and time by using LC/ and QC procedures, it is important to have a “long-term ref- HRMS-based untargeted metabolomics [115]. Inter-batch erence (LTR)” QC sample, as proposed by Dunn et al. [61]. In normalization was achieved using a pooled QC sample. Table 2 Selected publications dealing with recommendations and guidelines regarding metabolomics workflows References Experimental Sample collection Sample Data acquisition Data pre-processing Data processing Metabolite identification design preparation Sumner et al., – – Protocol and Instrumental conditions and Peak detection/integration – Metabolite identification 2007 [105] extraction performance, method methods validation Goodacre et al., – – – – Peak detection/integration Data mining, statistical – 2007 [197] analyses Dunn et al., QA/QC, large Large-scale studies Serum, plasma GC and LC-MS, samples, Data preprocessing Data processing workflow Levels of confidence, unknown 2011 [61] cohorts and pooled QC workflow metabolites Dudzik et al., QA Plasma, serum, Plasma, serum, Instrumental conditions, QA/QC QA/QC – 2018 [59] urine, cells, urine cells, batch and matrix effects, tissues tissues carryover Kirwan et al., Project Plasma, serum, – – – – – 2018 [57] planning urine, feces, saliva, CSF, tissues Broadhurst – – Pooled-QC Pooled-QC: precision, – Pooled-QC: inter batch – et al., 2018 preparation conditioning correction [98] González-Riano – – Plasma, serum, Multi-targeted metabolomics, Peak detection/integration Data cleaning normalization, GC-EI-MS (commercial or in-house et al., 2020 urine, feces, GC-MS, CE-MS, IMS, confounding factors, spectral libraries), LC-MS and [62] cells, tissues chiral analysis variable selection CE-MS solutions Rampler et al., – – Discussion on Absolute quantification, Peak detection /integration MS-based multi-omics, Metabolite and lipid annotation 2021 [30] protocols reference material merging metabolomics and lipidomics Castelli F.A. et al. Metabolomics for personalized medicine: the input of analytical chemistry from biomarker discovery to... They concluded that DBS and DUS stored at + 21 °C are 124]. Then, targeted detection of metabolites that are present stable for up to 4 weeks but are not stable over a 1-year period, in laboratory spectral databases can be carried out using peak whereas they showed good stability when stored at − 20 °C for integration software from instrument suppliers. The batch fu- 1 year. sion process would then be carried out more easily on the basis These two studies are based on the assumption that QC sam- of known metabolites (targeted data treatment) and no longer ples used for inter-batch normalization are stable in “reference based on m/z ratios and retention times alone and in a more storage conditions” (i.e., at − 80 °C or − 20 °C). However, some blinded way (untargeted data treatment). studies using absolute quantification and/or a calibration system highlighted some altered metabolite concentrations in plasma Studies dealing with large cohorts and requiring several lab- samples stored at − 80 °C up to five years. By using the oratories Although several laboratories with complementary Biocrates AbsoluteIDQ p180 targeted-metabolomics assay, technological expertise may be required in order to achieve Haid et al. observed significantly changed levels of amino acids, the largest metabolome coverage, most studies dealing with a acylcarnitines, glycerophospholipids, sphingomyelins, and the multi-platform approach are actually multi-omic studies in- sum of hexoses, with average increases or decreases of + volving a single metabolomics platform. In this case, general- 13.7% or − 14.5%, respectively [116]. Otherwise, Wagner- ly, each type of omics data is analyzed separately in order to Golbs et al. analyzed EDTA plasma samples stored for up to achieve molecular signatures, which are in turn collated and 16 years by gas and liquid chromatography-tandem mass integrated by using molecular network analysis and visualiza- spectrometry-based quantitative metabolomics. They found that tion software tools [125]. 226 out of 231 metabolites remained stable during the first seven Most of published metabolomics studies involving several years of storage [117]. laboratories or platforms with inter-laboratory comparison There is a limited number of large-scale studies using studies are dedicated to the evaluation of the performance untargeted metabolomics that have been published. Among and comparability of analytical methods [126–128]. those, a representative one is that of Dunn et al. who have The metabo-ring initiative gathered 5 NMR and 11 differ- reported on the molecular phenotyping of 1200 “healthy adults” ent LC/HRMS platforms with the objective of assessing the from the UK in the age range of 19–81 years, by using GC/MS reliability of untargeted metabolomics approaches in and LC/HRMS-based metabolomics [118]. The data were ac- obtaining comparable metabolomics profiles. Biological sam- quired across 11 months in 10 batches including samples from ples obtained from 2 different conditions were analyzed by the 120 subjects analyzed within a five-day period. Data were proc- partners using their own in-house protocols. It was observed essed using dedicated GC- and LC-HRMS workflows, and that, despite large differences in the number of spectral fea- inter-batch normalization was achieved thanks to a unique QC tures produced after post-processing and the heterogeneity of sample and the LOESS algorithm. From 259, 7813, and 7914 the analytical conditions and the data treatment, the spectral metabolite features initially detected in GC-MS, LC-HRMS information within and across technologies (NMR vs. LCMS) (positive mode), and LC-HRMS (negative mode), the imple- was highly convergent regarding 2 test datasets in terms of mentation of signal correction, batch integration, and quality statistical analysis [126]. assurance procedures led to 126, 2181, and 2283 metabolite Izumi et al. performed an inter-laboratory comparison features combined within a single multi-batch data matrix and study on cell line extracts including 12 participating laborato- available for statistical analyses. ries using their own analytical methods (capillary electropho- Another interesting study is that of Sindelar et al. on the use resis coupled to mass spectrometry, GC/MS and LC/MS with of metabolomics to highlight prognostic markers of COVID- different kinds of high-resolution mass spectrometers and 19 severity [119]. Seven hundred and four human plasma chromatographic conditions, corresponding to 15 and 9 samples were collected at six-month longitudinal points from methods for hydrophilic and lipophilic compounds, respec- 341 patients, and SRM 1950 was used as QC sample. Given tively). The aim of the study was to evaluate issues in inte- that the metabolic profiles were acquired over several months, grating different kinds of metabolomics data. Overall, 203 the combined data showed strong batch effects that proved metabolites and 580 lipid species were detected by at least efficiently corrected by combined batch correction [120]. one analytical method, among which 148 hydrophilic metab- Chromatographic retention time shifts are the main issues olites and 285 hydrophobic metabolites were detected by at regarding batch fusion over long periods. Particular software least two methods [127]. solutions have been described to correct within- and between- Finally, in their article, Yu et al. describe the Consortium of batch variability drifts in terms of mass accuracy, intensity, Metabolomics Studies, which was established in 2014 for fos- and retention times [121, 122]. Nevertheless, if there are too tering large-scale collaborative research on medical cohorts large differences in retention times between batches, it can and epidemiology. This initiative includes 47 cohorts from become difficult to correctly align peaks with common auto- Asia, Europe, North America, and South America, and blood matic signal detection and alignment software tools [123, samples were analyzed by 17 platforms. It appeared, from 2 Castelli F.A. et al. feasibility studies, that the overlap between any 2 different small set of compounds and thus suffer from such a limited laboratories in terms of detected metabolites ranged from 6 metabolite coverage. New MS/MS workflows as implement- to 121 metabolites within 5 leading laboratories, and that the ed on high-resolution mass spectrometers such as Q-TOF or median Spearman correlation coefficient was of 0.79 on 111 Orbitrap-based instruments have merged as powerful alterna- metabolites detected by two platforms. Absolute concentra- tive strategies [131]. Such acquisition methods referred to as tions were provided on only 31 metabolites across the 5 plat- data independent acquisition (DIA) including the SWATH forms, and 28% of identified metabolites were not listed in (accounting for Sequential Window Acquisition of all public databases [128]. Theoretical fragment-ion spectra) approach are being increas- All these studies highlighted a modest overlap in terms of ingly used in metabolomics by enabling simultaneous metab- metabolite detection from one facility to the other, and two of olite identification and quantification through the acquisition them reported on inconsistencies in compound identifiers that of MS/MS spectra for all analytes in a single run [131, 132]. limit the integration of datasets [127, 128]. This is especially Although slightly less sensitive than traditional LC-HRMS or the case with isomers that can be discriminated or not from MRM approaches, recent publications highlighted the poten- one platform to the other. Such issues could be addressed by tial of SWATH approaches for unambiguous compound de- the development and sharing of spectral databases [33], by tection and accurate quantification in complex samples fostering the evaluation and the use of standard reference ma- [133–136]. With constant instrumental improvements (e.g., terials within metabolomics data production facilities [127, data acquisition speed on Orbitrap mass analyzers, in collision 128], and of course by providing absolute concentrations on energies management, sensitivity, MS2 spectra identified metabolites present in the chemical libraries of data deconvolution), one can imagine great potential of such ap- production facilities, in the frame of untargeted approaches. proach for future biological metabolomics applications. As a sort of proof of principle, Zha et al. recently reported on an Analytical validation of biomarker candidates innovative SWATH-to-MRM approach. Thus, a high- highlighted by untargeted metabolomics: the need coverage targeted metabolomics method with 1303 metabo- for targeted quantitative metabolomics approaches lites in one injection was developed to profile colorectal tis- sues [137]. The success of such an approach would make LC- As mentioned above, confident metabolite identification is HRMS-based metabolomics both a screening and quantitative required for biomarker discovery. As a corollary, quantitative confirmatory technology. information about measured metabolites instead of relative differences must be provided if the biomarkers are to be used in clinical settings for diagnostic purposes through the defini- How to transfer metabolomics signatures tion of normal ranges of metabolite concentration. In addition from the research laboratories to the field to providing relevant data for answering clinical questions, metabolite concentrations expressed as molarity units will also Metabolomics signatures are generally defined by clinical re- facilitate access to the dynamics of the metabolome and of search studies, which cannot be considered as routine care course the integration of metabolomics data with those of practice. They are complex for several reasons. First, they other omics, while also enabling facile comparison of results are multi-parametric in that they often consist of few tens of among laboratories and studies. metabolites. They often deal with small concentration varia- Accurate and validated metabolite quantification is gener- tions from one group to the others, with concentration ratios ally accomplished by using targeted LC-MS/MS-based ap- often less than a factor five. Some of the metabolites of interest proaches using low-resolution triple quadrupole instruments may be only partially characterized (i.e., for example a com- operating in the multiple reaction monitoring mode (MRM) pound class, such as an acylcarnitine species including a hy- and using metabolites labeled with stable isotopes (e.g., 13C, droxylated carboxylic acid in which the position of the hy- 15 N, 2H) [129]. Such an approach can also be used to validate droxyl group cannot be precisely located). Furthermore, when the results obtained by untargeted LC-HRMS metabolomics. obtained by using untargeted approaches, results are not For example, the concentrations of 4 tryptophan metabolites expressed in molarity units but rather as peak areas, which have been recently measured by LC-MS/MS (multiple reac- obviously limits data reuse and sharing. Finally, these molec- tion monitoring—MRM—mode with isotope dilution) in the ular signatures are often not validated due to a lack of valida- serum of 218 patients with acute decompensation and ACLF tion cohorts, and also of specificity studies which are time in cirrhosis, and demonstrated excellent correlation with the consuming, expensive, and difficult to design. corresponding MS signals observed under LC-HRMS condi- Consequently, metabolomics signatures obtained from tions [130] (Fig. 4). untargeted metabolomics cannot be directly used in the routine Although targeted LC-MS/MS (MRM) methods offer the care practice. There is a need for simplification and for moving best quantification sensitivity, they are often focused on a to quantitative results. Simplification can be achieved through Metabolomics for personalized medicine: the input of analytical chemistry from biomarker discovery to... Fig. 4 Comparison of metabolite concentrations measured by using with different levels of cirrhosis decompensation. Correlation analyses untargeted and targeted approaches. Correlation between LC-HRMS data were achieved by calculating Pearson correlation coefficients (r). (peak area, exactive instrument) and absolute quantification data (ng/mL, Experimental conditions are displayed in the publication of Claria et al. Waters Xevo TQ-XS instrument) obtained for tryptophan, quinolinic ac- [130] id, kynurenine, and kynurenic acid measured in the serum of 217 patients statistical analysis tools enabling to select a small number of key operations and (ii) outputs of metabolomics workflows are com- components of the signatures. This is, for example, the case with plex molecular signatures of few tens of metabolites, often with the biosigner algorithm, which enables to find the smallest fea- small concentration variations. In the field of personalized med- ture subset which significantly contributes to the performance of icine, the challenges will then lie in the longer term in the devel- a multivariate statistical analysis model [138]. Regarding quan- opment and use of field rapid diagnostic tests based on the de- tification, multiplexed targeted assays can be developed and velopment of biosensors for the multiplexed and quantitative implemented for specifically monitoring essential metabolites detection of several biomarker candidates. obtained from complex molecular signatures in order to confirm results obtained from untargeted metabolomics experiments, as already described in the previous section. Metabolomics at medical laboratories Furthermore, many actors and structures are involved in healthcare systems: clinical units in hospitals, medical laborato- The clinical context has a strong impact on the manner of ries in hospital settings or outside the hospitals, physician’s of- translating metabolomics signatures into clinical practice. It fices, pharmacists, and at least the patient at home. Current lim- can be intended for a critical care, a chronic disease follow- itations in terms of data reuse and interoperability combined with up, or a genetic rare disease. Modalities such as the frequency a lack of validation data for the molecular signatures generated of analysis, the delivery time of the results, the geographical make metabolomics not yet easily usable in routine care. availability (local, regional) of equipment, and the routine However, the main users in the short term could be clinical workflow have to be defined by taking into account this con- biologists and chemists, as well as non-experts, working in med- text. These aspects are usually managed in a clinical laborato- ical laboratories in a hospital context, near the clinical units. It is ry by a clinical biologist. One role of the biologist is to inte- likely that in the near future the use of MS-based approaches will grate these requirements with analytical constraints, by setting remain limited to hospital settings, operated by trained staff. This a framework in which the sample pathway, the analytical because (i) the sample preparation requires technically advanced quality, and the delivery of the results are well defined. Castelli F.A. et al. Delivering a metabolic signature to a physician is challenging. as bile acids [141], formic acid [142], oxalic acid [143], or This means to translate complex and numerous data into a self- sialic acid [144]. Such assays are mainly based on the moni- explanatory analysis report, which must be available and inter- toring of enzymatic cofactors such as NADH or NADPH that pretable for any clinician specialized in a given medical area. are consumed by the enzymatic reaction together with the This translation requires processing, integration, and interpreta- analyte. Although simple to implement when kits are commer- tion of data, in order to transmit a suitable information to the cially available, these approaches may suffer from low spec- clinical issue. Clinical biologists will have a key role in the results ificity and may lead to underestimated values, as recently re- transmission by integrating the biological message into the global ported for bile acids [145]. They are also prone to interfer- clinical context. This requires rethinking medical biology by no ences, as observed with oxalic acid for which vitamin C inter- longer reasoning in terms of isolated markers reflecting an organ- feres [143]. Furthermore, as enzymes may process all the ic function, but rather in terms of molecular signatures reflecting members of a chemical class, some enzyme assays enable the stage of a disease at a given time. This is a major paradigm the determination of total concentrations rather than individual shift from a focus on broad categories of disease, to a more ones, as observed for bile acids and triglycerides, for example. holistic approach that will integrate a patient’s metabolic status, Finally, such assays suffer from low multiplexing capabilities. impacted by all of their co-morbidities and their environment. To For all these reasons, if enzymatic methods are useful to achieve this, clinical biologists will have to upgrade analytical analyze a metabolite or possibly a set of metabolites belonging technologies to generate data, integrate bioinformatics solutions, to a given chemical class, it seems unlikely to use them for and develop an automated algorithm to express targeted and more complex metabolomics signatures as they require the accurate results from complex data. Actually, clinical biochem- ability to find and produce dedicated enzymes and to set up istry is likely to undergo the same technological revolution as conditions allowing enzymatic activity measurement in differ- molecular genetics has undergone over the last 20 years with the ent types of biological media. advent of genomics and next-generation sequencing. One of the main challenges will be to obtain concentrations of key metabo- Laboratory immunoassays for the detection of lites expressed in units of molarity, in order to meet the short- metabolites in biological fluids term needs of clinicians, while at the same time having the pos- sibility of building databases of metabolic profiles that will be As main tools of immunoassays, antibodies enable the detec- interrogated in different medical contexts. tion and quantification of specific biomarkers and are partic- Although there are some liquid chromatography coupled to ularly suitable for molecules with a molecular weight above mass spectrometry systems that are CE-IVD (i.e., European 3000 Da [146]. Performances of an immunoassay will rely directive for In Vitro Diagnostics) approved for clinical bio- first on the affinity and specificity of the antibodies used: chemistry, toxicology, or therapeutic drug monitoring, and polyclonal antibodies (i.e., collection of purified immuno- mass spectrometers can be used in operating rooms [139, globulin molecules obtained from immunization of animals 140], not all medical laboratories are and will be equipped such as rabbits, sheep, donkeys, or goats), usually easier and with mass spectrometers (especially high-resolution instru- faster to produce than monoclonal antibodies (i.e., single class ments) and very few of them with nuclear magnetic resonance of antibodies produced by a monoclonal immortalized B lym- instruments. Thus, other analytical methods, such as enzyme phocyte), generally display high affinities for their target but assays, immunoassays, and biosensors, have to be envisaged limited specificity. Monoclonal antibodies (mAb) are pre- for metabolite detection. For example, enzyme- and immuno- ferred tools for the development of accurate and specific im- assays are already widely and routinely used in hospitals for munological tests. Second, performances will depend on the clinical biochemistry and therapeutic drug monitoring, as part detection system that can be achieved using various methods, of industrial automated in vitro diagnostic systems. including among others isotopic labeling (radioimmunoas- says), enzyme reaction with UV-visible or fluorescence or chemiluminescence detection, or colloidal gold particles. Alternative tools to mass spectrometry There are two main immunoassay formats: competitive and and nuclear magnetic resonance instruments immunometric, depending on the size of the analyte. In for making metabolomics valuable in clinical immunometric assays, a first capture antibody, specific for laboratories the antigen, is bound to a solid surface. The antigen is then added, followed by addition of a detection antibody. The latter Enzyme assays for monitoring metabolites in binds the antigen to a different epitope from the capture anti- biological fluids body. Thus, this assay format is adapted to analytes containing a least two epitopes (i.e., molecules having a molecular mass As previously emphasized, enzyme assays are already widely above 1000 Da, assuming that an epitope includes at least 5 used in clinical laboratories for monitoring metabolites, such amino acids). Conversely, in competitive immunoassays, the Metabolomics for personalized medicine: the input of analytical chemistry from biomarker discovery to... antibody is immobilized on a support and the detection is measurements are achieved in a liquid phase), such as EMIT achieved through a labeled antigen. The addition of the sam- (Enzyme Multiplied Immunoassay Technique; see Fig. 5a), ple containing free antigen induces an antibody binding equi- CEDIA (Clone Enzyme Donor Immunoassay), FPIA librium between the free antigens and the labeled ones. This (Fluorescence Polarisation Immunoanalysis), and KIMS assay format is preferably used when only one antibody is (Kinetic Interaction of Microparticle in Solution) [153]. available, or when the analyte has only one epitope (i.e., small Laboratory immunoassays can also take place in heteroge- antigen < 1000 Da), which is the case for metabolites. neous phases. In this case, the assays are performed in several Before the development and popularization of LC-MS ap- steps with reagents added and optionally washed or separated proaches in the 1990s, competitive immunoassays were wide- at the site of the antigen/mAb complexes. Enzyme linked ly used for the sensitive detection of drugs and their metabo- immunosorbent assays (ELISA) are probably the most used lites for pharmacokinetics and therapeutic drug monitoring, heterogeneous phase assay formats for the detection and quan- and for clinical chemistry, as it was for example the case for tification of biomarkers in biological media. Since it requires steroid hormones [147, 148]. Nowadays, competitive immu- multiple steps (and washing steps in between), ELISA (Fig. noassays are still used for a number of clinical biology analy- 5b) needs to be performed by trained staff with laboratory ses such as the detection of 25-hydroxyvitamin D (vitamin D equipment. metabolite) by radioimmunoassay [149], the detection of ox- Finally, other types of assays aiming at overcoming the idative stress biomarkers based on the combination of limitations of competitive formats have been specifically de- microfluidics and fluorescent immunoassay [150], and the veloped for the detection and quantification of small mole- detection of hormones such as progesterone by coupling mi- cules. Some of them, such as SPIE-IA (solid-phase crochip electrophoresis and chemiluminescent immunoassay immobilized epitope-immunoassay), AIA-NIA (anti-idiotypic [151] or cortisol using a paper-based immunosensor with a antibody-based non-competitive immunoassay), AICA-NIA colloidal-gold labeled immunoassay [152]. (anti-immune complex antibody-based non-competitive im- Various types of competitive assays have thus been devel- munoassay), and OS-NIA (open sandwich non-competitive oped and implemented on automated in vitro diagnostic de- immunoassay), are described in Fig. 6 [155]. vices. The most popular are competitive immunoassays in Although some of these assay formats are attractive and homogeneous phase (i.e., reagents, samples, and inspiring for the design of biosensors for personalized Fig. 5 Competitive laboratory immunoassays for small molecules. a Principle of EMIT (Enzyme Multiplied Immunoassay Technique) [153]. EMIT is a competitive immunoassay in homogenous phase in which an analyte analog is bound to an enzyme using nicotinamide-adenine- dinucleotide (NAD) as a cofactor. The enzymatic reaction generates NADH which is detected by spectrophotometry at 340 nm. A competition between the analyte and the enzyme bound analog takes place toward the antibody. The amount of NADH produced is directly related to the amount of analyte present in the sample. b Competitive ELISA [154]: Antibodies are immobilized on the solid support. A competition takes place between an analyte analog coupled to an enzyme and the free analyte in the sample. The detection is achieved through en- zymatic activity Castelli F.A. et al. medicine purposes, immunoassays for small molecules suffer such as cortisone or many synthetic steroids such as prednis- from several limitations. Indeed, the lack of sensitivity and olone or prednisone [156]. The challenge to develop assays specificity of antibodies for small molecules are their main based on the recognition of metabolites by antibodies firstly weaknesses. Although many antibodies have been developed lies in the possibility to derive these small chemical molecules, against small molecules, they can suffer from cross-reactions. to make them haptens (i.e., to link them covalently to carrier This is the case, for example, for antibodies against cortisol proteins whose role is to provoke/favor the immune response which have cross-reactivity with many endogenous steroids in immunized animals, naturally not immunogenic). In some Fig. 6 Non-competitive laboratory immunoassays for small molecules signal strength is proportional to the amount of Ab2α labeled and bound [155]. a SPIE-IA (solid-phase immobilized epitope-immunoassay): This to the Ab1/antigen complex. c AICA-NIA (anti-immune complex format is based on the use of a single antibody that acts as both capture antibody-based non-competitive immunoassay): This assay uses an and detection antibody. It takes place in four steps: (i) Analytes are cap- immobilized (Ab1) and an anti-metatypic (Ab2) antibody, the latter sta- tured by immobilized antibodies. (ii) Analytes are covalently bound to the bilizing the antibody/analyte complex. It takes place in two stages: (i) The immobilized antibodies with the help of a reagent (e.g., glutaraldehyde, analyte binds to the Ab1. (ii) Ab2 is added and binds the analyte-antibody carbaonyldiimidazole). (iii) C-Analytes are then released from the complexes. The intensity reflects the amount of Ab2 that has bound. d immobilized antibodies by denaturation with a solvent. (iv) Detection OS-NIA (open sandwich non-competitive immunoassay): This format is antibodies coupled to an enzyme can then fix the analytes. b AIA-NIA based on the association of free VH and VL chains from the variable (anti-idiotypic antibody-based non-competitive immunoassay): This for- domain of an antibody, which dissociate in the absence of the antigen mat requires the use of three antibodies: an immobilized primary antibody (i.e., the analyte). It takes place in two stages: (i) The VL chains, conju- (Ab1), an anti-idiotypic antibody (Ab2α), and a labeled anti-idiotypic gated to a carrier protein, are fixed by immobilized antibodies. (ii) The antibody (Ab2β) and is performed in four steps : (i) The analyte binds analyte and the labeled VH chains are added. The binding of the antigens to Ab1. (ii) Ab2β is added to block the remaining Ab1 free binding sites. to the VL chains allows the association of the VL and VH chains. The (iii) Ab2α are then added to capture only the Ab1/analyte complexes intensity of the signal is proportional to the quantity of labeled VH chains (Ab2β/Ab1 complexes cannot be captured due to steric hindrance). The present Metabolomics for personalized medicine: the input of analytical chemistry from biomarker discovery to... cases, complex coupling chemistry procedures are necessary affine antibody is difficult as the molecule is small [146]. to obtain an immunogen [157]. Furthermore, antibodies of Secondly, the decreased signal observed in competitive interest have to be selected on their specificity to recognize methods is more difficult to interpret than the appearance of exclusively the molecule, without recognition of the derived a signal [146]. Moreover, these tests are neither suitable for molecule, or of related molecules. The smaller the molecule, multiplexed detection nor for field quantification, and are sub- the more difficult this selection is to achieve. Competitive ject to matrix effects. Indeed, in the case of small endogenous immunoassays also exhibit poorer sensitivity than molecules, the challenge is important: some of them may have immunometric ones because of the use of a limited amount relatively small variations (less than a factor of 5) between a of reagent, which does not facilitate the formation of antigen/ physiological and a pathological state, they generally require a antibody complexes. preliminary sample extraction step, and only the integration of Finally, there could be issues of loss of performance, po- data from several biomarkers will allow the interpretation of a tentially in terms of specificity, when moving from mass biological signature. The test must be accurate, robust (with spectrometry–based methods to immunoassays. In this con- little matrix effect), quantitative, highly multiplexable, and text, there are many studies aimed at comparing the perfor- easily interpretable, and must integrate automated sample mances of immunoassays and mass spectrometry–based ap- preparation before analysis. Therefore, the PoC tests for fine, proaches. Most of them deal with drugs and their metabolites, multiplexed, integrated quantification of multiple metabolites and, to a lesser extent, with endogenous metabolites such as of different natures must address unmet needs and will require steroids, thyroid hormones, and vitamin D derivatives. Some a large research effort which will be driven by the need of of these studies show that these two methods exhibit the same clinicians [167]. kind of performances [158–160], whereas others point out Research on rapid tests has developed considerably, espe- variable quantitative performances between immunoassays cially regarding nanoparticles-based devices [168] and [161], and specificity issue with over- or underestimated con- microfluidics [169], paving the way to the emergence of centrations measured in immunoassays [162–165]. new diagnostic devices [170]. Wearable chemical sensors en- Anyway, despite these limitations, it is important to con- abling real-time on-body analytical chemistry are in increas- tinue working on new small-molecule immunoassay formats ing development. However, they only allow monitoring elec- given their sensitivity, their relative simplicity of implemen- trolytes and few metabolites such as glucose or lactic acid at tation, and the lack of efficient alternative methods, when the time of writing this review [171]. methods based on mass spectrometry and proton nuclear mag- Biosensors are devices that combine the physical properties netic resonance spectroscopy cannot be used. of a transducing platform (electrical, optical, etc.) with those of a biological or synthetic component in order to detect, i.e., Biosensors for small molecules: toward point of care sense, the presence and, possibly, quantify the concentration tests for metabolomics? of a target compound, i.e., the analyte. This combination, if properly addressed, benefits from both the natural selectivity Point-of-care (PoC) tests are medical laboratory diagnostic of a wide range of bioreceptors (enzymes, antibodies but also tests intended to be carried out in the direct proximity of the nucleic acids and even viruses, bacteria, and whole cells) or patient, at the physician’s office, in pharmacies, in medical synthetic molecules (i.e., molecularly imprinted polymers centers, in the emergency rooms of hospitals, or even in pro- (MIPs), aptamers, etc.) and the sensitivity of the devices fessional laboratories. Whether in hospitals or not, PoC tests converting, i.e., transducing, the binding event between the are expected to give results within a short period of time (30 bioreceptors and the analyte in an analytical signal. min). Those tests thus require the smallest possible logistical Depending on the nature of this signal, several transduction footprint, in terms of transport and space and storage condi- methods could be employed, such as optical, electrochemical, tions, as well as handling time, and are designed to be carried thermal, or piezoelectric biosensors. out by staff not necessarily trained in laboratory medicine Biosensors are one of the most important representatives of (nurse, medical assistant), or even by the patient himself or PoC devices that are having a remarkable impact on patients’ his relatives. Lateral flow immunoassay (LFIA) is probably life and, thus, are also being pursued with the highest interest the ideal on-field point-of-care test since it does not require by the research community. Indeed, PoC biosensors’ charac- complex sample preparation or further steps after the deposi- teristics should reach the criteria “REASSURED,” i.e., real- tion of the sample and can be performed easily by untrained time connectivity, ease of specimen collection and environ- staff. Although mainly used for protein targets, competitive mental friendliness, affordable, sensitive, specific, user friend- LFIA, whose principle is described on Fig. 7, have been de- ly, rapid, equipment free, delivered [172], as they should pro- veloped for small molecules. However, they are more com- vide real-time connectivity and easy specimen collection. plex to implement and less sensitive than non-competitive Nanotechnology can push even further the characteristics LFIA for different reasons. Firstly, obtaining a specific and of PoC biosensors by introducing nanomaterials, guaranteeing Castelli F.A. et al. Fig. 7 Principle of competitive lateral flow immunoassays: The device is excess sample. In the absence of the target analyte (negative sample), composed of four parts: (i) a sample pad, on which the sample is depos- labeled analog analytes move through the strip and bind on both test ited; (ii) a conjugate dried buffer, containing the labeled analyte analog; and control lines. In a positive sample, a competition takes place on the (iii) a nitrocellulose membrane, on which are found test line(s) composed test line between the analyte and its labeled analog. As before, the excess of antibodies recognizing the analyte(s), and control line(s) formed by of labeled analyte analog is captured by antibodies in the control line. antibodies which recognize the labeled analyte analog; and (iv) an absor- Thus, a signal is only observed on the control line. Adapted from refer- bent paper, which serves to pump the liquid sample and reserves any ence [166] numerous advantages, such as ultra-high sensitivity, further Theoretically, these methods allow the production of big miniaturization if necessary, nanoreceptors with high stability amounts of aptamers at low cost and take less time compared and specificity, and innovative detection mechanisms, im- to antibody production, as well as avoiding the use of animal proving the performances while decreasing the associated testing and batch-to-batch variations. Of course, this allows costs. the production of aptamers for toxic molecules that would kill Most of the challenges of small-molecule detection the animal in the immunization process. Moreover, stability are directly related to the use of biorecognition ele- and affinity toward the target can be modulated by modifying ments. As previously emphasized, due to the low mo- the sequence of the aptamer, along with adding other elements lecular weight and simple structure of small molecules, to the sequence depending on its intended use (fluorophores, they show a lack of immunogenicity when it comes to biotin, thiol groups, etc.), which is especially interesting for antibody production. For this reason, there have been assays combining aptamers with nanomaterials. As SELEX is strong efforts to develop other types of bioreceptors in performed by incorporating negative counter-selection order to have better performance with this type of mol- rounds, a high specificity and selectivity are ensured. In con- ecules. Among them are aptamers and MIPs, which can trast, antibodies often show cross-reactivity between similar be successfully applied to PoC nanobiosensors. molecules [178]. Aptamers are nucleic acid (DNA or RNA) short sequences Aptamers are approximately one tenth of the size of an that bind to a specific target and are determined through se- antibody, with enough recognition surface area to target small quential evolution of ligands by exponential enrichment molecules, unlike antibodies, which rely on epitopes that are (SELEX) [173, 174]. A pool of random nucleic acid se- often scarce in small molecules [179]. Compared with anti- quences is repeatedly incubated with the target analyte, and bodies, the small size of aptamers also allows a high immobi- the unbounded sequences are discarded. After various rounds lization rate at the surface of a biosensor, increasing its sensi- with different stringencies, the sequences with higher affinity tivity. Besides, they can be regenerated for long-term moni- for the target are selected. The SELEX methodology has been toring and are not susceptible to irreversible denaturation, in successful in developing aptamers for small molecules. This is contrast to antibodies. Moreover, aptamers have a long in part thanks to the flexibility of the method, which can be shelf-life and stability under strong conditions such as adapted to the nature and needs of the target molecule (cap- temperature, chemicals or pH, which makes its transpor- ture-SELEX, capillary-SELEX, or nitrocellulose-SELEX, for tation, storage, and performance much easier in remote instance). [175, 176]. Conventional and capture-SELEX have areas [178]. As depicted in Table 3, different platforms been reported as suitable options for small molecules [177]. and detection strategies have been successfully and Metabolomics for personalized medicine: the input of analytical chemistry from biomarker discovery to... sensitively detecting small molecules in biological sam- a precise combination of one or several functional monomers ples by using aptamers in the last decade. For instance, and the desired target, i.e., “template molecule,” along with microfluidic devices, lateral flow and electrochemical other components present in the polymerization solution, such systems are some of the potential platforms where as a cross-linker molecule. The utmost important challenge aptamers can be implemented (see Fig. 8). that needs to be addressed in the first stages of the develop- To summarize, aptamers have shown potentially great ad- ment of a MIP would be to study the binding interactions vantages over the use of antibodies for diagnostic applications among the monomers and between the monomer and the tem- since their discovery in 1990. This is especially of interest for plate, which must be guaranteed in order to form a spontane- small molecules for which antibodies are not well suited. ous and stable template-monomer complexation. From this, However, their implementation into PoC nanobiosensors is several other parameters must be considered for the formation rather scarce [169], and despite promising results, there is still of a single-molecule MIP or a “class-selective” one, which is no commercialized diagnostic tests based on aptamers. particularly designed to detect a family of related molecules. Studies are needed to evaluate their performance in biological Some of them are the template-to-monomer (T:M) ratio, the media for diagnostic tests involving metabolites. crosslinking degree, porogenic solvent selection, and some For the last decades, it has been a challenging task to gen- physical parameters, such as temperature, pH, and agitation erate polymeric matrices with homogeneously distributed and of the polymer solution. highly specific cavities, thoroughly designed to match the di- The use of MIPs as another metabolomics tool has been mensions and chemical functionality of a target molecule. considered an interesting approach due to their fast produc- This type of bioreceptors is known as molecularly imprinted tion, robustness, chemical inertness, cost-effectiveness, and polymers (MIPs), which are considered synthetic long-term stability. If precisely tuned, they can withstand ex- biorecognition elements resulting from the polymerization of treme values of pH and temperature [180, 181]. Unlike current Table 3 Characteristics and performance of the main biosensors currently available for small-molecule detection in biological fluids Analyte Transducer Bioreceptor LOD Response Biological Ref. time fluid Cocaine and synthetic Colorimetric Aptamer 10 μM 5 min Saliva and Luo et al., 2019 cathinones urine [198] Cocaine Microfluidic, Aptamer 10 μM 1–2 min Blood Swensen et al., electrochemistry serum 2009 [199] Adenosine trihosphate (ATP) Lateral-flow assay Self-assembly of split 2 μM 10 min Blood Chen et al., 2012 aptamers fragments serum [200] Adenosine Electrochemistry uPAD Aptamers 5.7 μM 10 min Urine Fu et al., 2017 [201] Cocaine, ATP Fluorescence Exonuclease-mediated 500 nM 25 min Urine Canoura et al., aptamer digestion 2018 [202] Tetrahydrocannabinol Magnetoresistive sensor Antibodies competitive 10 ng/ml < 15 min Saliva Lee et al., 2016 detection [203] Ochratoxin A, aflatoxin B1, Localized surface Aptamers on gold 0.56, 0.63, 0.87, 30–60 min Serum Park et al., 2017 ATP, potassium ions plasmon resonance nanorods 1.05 pM [204] L-Tyrosinamide Fluorescence Aptamer 200 nM < 10 min Urine Ruta et al., 2009 polarization assay [205] Phenytoin CMOS BioMEMS 4.06 μg/ml 25 min Artificial Yen et al., 2020 samples [206] Dopamine, cortisol, serotonin Thermal variation MIPs and thermal 8 μM / Serum and Diliën et al., transducers urine 2017 [186] Carnitine Potentiometric MIPs, Radical 80 μM / Urine Moret et al., polymerization 2014 [187] Dopamine Ratiometric MIPs and nanoporous Au 0.1 μM 2 min Artificial Yang et al., 2019 electrochemical CSF [207] Glucose Electrochemial MIPs and AuNPs 1.25 nM 30 min Serum Sehit et al., 2020 [188] Castelli F.A. et al. Fig. 8 Overview of the possible adaptation of aptamers in different using an electrochemical aptamer-based sensor suitable for point-of- detection platforms combined with nanomaterials. (a) Gold nanocap- care applications extracted from reference [210]. (c and d) Aptamer- supported up-conversion nanoparticles for fabrication of a solid-phase based lateral flow test strip for rapid detection of zearalenone in corn aptasensor for ochratoxin A detection. Extracted from reference [209]. samples. Adapted from reference [211] (b) Calibration-free measurement of phenylalanine levels in the blood and typical immunoaffinity-based approaches, MIPs have In these cases, the integration of nanomaterials clearly up- demonstrated a higher sample load capacity for small mole- grades the sensitivity, either by signal amplification or better cules (MW below 3 kDa), resulting in higher recoveries for immobilization of the biorecognition elements [190]. Another further analytical applications, as well as displaying a slightly inconvenience is the complexity of biological fluids, which higher selectivity and specificity toward smaller targets [182]. might contain contaminants causing cross-reactivity, or even Based on the final applicability of the desired sensor, sev- nucleases, that are detrimental for these bioreceptors [191]. eral approaches have been developed. For detecting tamoxi- While electrochemical platforms are well known for their ad- fen, for example, an estrogen receptor used to prevent breast vantages in complex matrices like blood (i.e., glucometer, the cancer, a successfully electropolymerized MIP composed of most-known example), optical platforms still face the problem o-phenylenedediamine and resorcinol was developed [183]. of interferences mainly due to the viscosity and optical prop- This methodology demonstrated a better diffusion rate, per- erties of the samples (e.g., the deep red of blood causes back- meability, and binding affinity to the target. Acetaminophen ground interference) [192] [193]. has also been investigated for medical and clinical purposes, by using signal enhancers, such as gold nanoparticles [184], or developing a thin MIP layer that enhances the selectivity Conclusion [185]. Other metabolites have also been studied, such as do- pamine [186], carnitine [187], and glucose [188], among Metabolomics workflows, including sample preparation, MS others, as it can be seen in Table 3 and Fig. 9. and/or NMR analyses, data pre-processing, statistical analy- Anyway, to our knowledge and as for aptamers, there are ses, and data visualization, have been developed since the still no commercialized diagnostic tests based on MIPs. These 2000s and have now reached a certain level of maturity. two types of synthetic bioreceptors (aptamers and MIPs) need Data production methods have gained in sensitivity and ver- to address the issue of properly working in real biological satility, suggesting the possibility of achieving metabolite de- fluids, as well as for enzyme- and immunoaffinity-based sys- tection, identification, and quantification at the same time. tems, as it was aforementioned [189]. Most metabolites are Many guidelines covering pre-analytical stages, data acquisi- often found in ultra-low concentrations in biological samples. tion, and study design have been published. Quality Metabolomics for personalized medicine: the input of analytical chemistry from biomarker discovery to... Fig. 9 MIP-based biosensing platforms. a Gold nanoparticle (AuNP) et al. [186]. c Schematic diagram of the synthesis process of MIPs/pThi/ decorated MIPs using o-PD and glucose as monomer and template mol- NPG electrodes, along with their respective DPV measurements and the ecules, respectively. CV and DPV measurement of each step. Extracted calibration curves for dopamine detection. Adapted from Yang et al. from Sehit et al. [188]. b Schematic representation of the heat flow [207] through the MIP and NIP-coated thermocouples. Extracted from Diliën management processes have been proposed, become consen- centralized data production platforms equipped with (high- sual, and are more and more used. Data warehouses dedicated resolution) MS or NMR instruments. However, not all medi- to metabolomics have been developed, improving data shar- cal laboratories are and will be equipped with these instru- ing. Despite this, no diagnostic test based on metabolomics ments. Thus, there is a need to consider alternative methods, has yet been marketed. The main issues are linked to lacks of such enzyme assays, immunoassays, and biosensors. In par- standardized data production tools and interoperability, to in- ticular, even if enzyme assays have been and are still widely appropriate design of clinical trials for the discovery and val- used for monitoring small molecules in biological fluids, they idation of metabolomics signatures, and to the difficulty of are restricted to few key metabolic intermediates present at integrating multiscale biological information to generate high concentrations, and they have low multiplexing capabil- knowledge and predictive models. ities. Immunoassays are also very popular in the field of clin- However, more than 600 articles dealing with metabolomics ical and environmental chemistry, regarding the detection of for the investigation of medical cohorts have been published drugs and more generally xenobiotics. However, they still over the 2015–2020 period, some of them leading to proposed suffer from limitations linked to the difficulty of generating metabolomics signatures of disease diagnosis and severity, and antibodies having satisfactory sensitivity and specificity, response or non-response to treatments. In this context, it is now which limits the design of biosensors for small molecules. time to consider how to be prepared to efficiently transfer future Aptamers and molecularly imprinted polymers are attractive metabolomics signatures to clinical settings. First of all, meta- and promising alternatives to antibodies, especially in the field bolomics signatures obtained from untargeted metabolomics of small molecules, but further research efforts are needed to cannot be directly used in the routine care practice. There is a evaluate their relevance in complex biological media. need for simplification and for moving to quantitative results. To conclude, it is important to intensify research in analyt- Indeed, many actors and structures are involved in healthcare ical chemistry, not only in the generation of metabolomics systems, such as clinical units in hospitals, medical laboratories data for producing interoperable and reusable data, but also in hospital settings or outside the hospitals, physician’s offices, in the field of point-of-care tests, in order to be ready when pharmacists, and at least the patient at home. molecular signatures can be used in routine care practice. A first step regarding the transfer of metabolomics signa- Furthermore, the small metabolite concentration variations tures to the field could rely on medical laboratories in hospital that are often observed between groups in many metabolomics settings. Clinical biologists will certainly have a key role in and lipidomics studies represent an important limitation for data production and interpretation, and for transmitting the clinical translation, which will call for new ways of thinking results, clinicians by integrating the biological message into in the fields of analytical chemistry and data sciences to over- the global clinical context. In this case, one can think about come this issue. Castelli F.A. et al. Funding Christian Moguet is the recipient of a PhD grant funded by 8. Deng P, Li X, Petriello MC, Wang C, Morris AJ, Hennig B. CEA. We acknowledge funding from the MICROB-PREDICT project. Application of metabolomics to characterize environmental pol- The MICROB-PREDICT project has received funding from the lutant toxicity and disease risks. Rev Environ Health. 2019;34: European Union’s Horizon 2020 research and innovation program under 251–9. https://doi.org/10.1515/reveh-2019-0030. grant agreement No. 825694. This reflects only the author’s view, and the 9. Salem MA, Perez de Souza L, Serag A, Fernie AR, Farag MA, European Commission is not responsible for any use that may be made of Ezzat SM, Alseekh S. Metabolomics in the context of plant natural the information it contains. products research: from sample preparation to metabolite analysis. Metabolites. 2020;10. https://doi.org/10.3390/metabo10010037. Declarations 10. Castro-Moretti FR, Gentzel IN, Mackey D, Alonso AP. Metabolomics as an emerging tool for the study of plant- pathogen interactions. Metabolites. 2020;10:37. https://doi.org/ Ethics approval Not applicable 10.3390/metabo10020052. 11. González-Peña D, Brennan L. Recent advances in the application Source of biological material Not applicable of metabolomics for nutrition and health. Annu Rev Food Sci Technol. 2019;10:479–519. https://doi.org/10.1146/annurev- Statement on animal welfare Not applicable food-032818-121715. 12. Rivera-Velez SM, Navas J, Villarino NF. Applying metabolomics Conflict of interest The authors declare no competing interests. to veterinary pharmacology and therapeutics. J Vet Pharmacol Ther. 2021. https://doi.org/10.1111/jvp.12961. Open Access This article is licensed under a Creative Commons 13. Hood L, Flores M. A personal view on systems medicine and the Attribution 4.0 International License, which permits use, sharing, adap- emergence of proactive P4 medicine: predictive, preventive, per- tation, distribution and reproduction in any medium or format, as long as sonalized, and participatory. New Biotechnol. 2012;29:613–24. you give appropriate credit to the original author(s) and the source, pro- https://doi.org/10.1016/j.nbt.2012.03.004. vide a link to the Creative Commons licence, and indicate if changes were 14. Sagner M, McNeil A, Puska P, Auffray C, Price ND, Hood L, made. The images or other third party material in this article are included Lavie CJ, Han Z-G, Chen Z, Brahmachari SK, McEwen BS, in the article's Creative Commons licence, unless indicated otherwise in a Soares MB, Balling R, Epel E, Arena R. The P4 health spectrum credit line to the material. If material is not included in the article's - a predictive, preventive, personalized and participatory continu- Creative Commons licence and your intended use is not permitted by um for promoting healthspan. Prog Cardiovasc Dis. 2017;59:506– statutory regulation or exceeds the permitted use, you will need to obtain 21. https://doi.org/10.1016/j.pcad.2016.08.002. permission directly from the copyright holder. To view a copy of this 15. Bossuyt PM. Where are all the new omics-based tests? Clin Chem. licence, visit http://creativecommons.org/licenses/by/4.0/. 2014;60:1256–7. https://doi.org/10.1373/clinchem.2014.223339. 16. Pinu FR, Goldansaz SA, Jaine J. Translational metabolomics: cur- rent challenges and future opportunities. Metabolites. 2019;9. https://doi.org/10.3390/metabo9060108. 17. Trifonova OP, Maslov DL, Balashova EE, Lokhov PG. Mass spectrometry-based metabolomics diagnostics - myth or reality? References Expert Rev Proteomics. 2021;18:7–12. https://doi.org/10.1080/ 14789450.2021.1893695. 1. Oliver SG, Winson MK, Kell DB, Baganz F. Systematic function- 18. Trivedi DK, Hollywood KA, Goodacre R. Metabolomics for the al analysis of the yeast genome. Trends Biotechnol. 1998;16:373– masses: The future of metabolomics in a personalized world. New 8. Horiz Transl Med. 2017;3:294–305. https://doi.org/10.1016/j. 2. Junot C, Fenaille F, Colsch B, Bécher F. High resolution mass nhtm.2017.06.001. spectrometry based techniques at the crossroads of metabolic 19. Zhang X, Li Q, Xu Z, Dou J. Mass spectrometry-based metabo- pathways. Mass Spectrom Rev. 2014;33:471–500. https://doi. lomics in health and medical science: a systematic review. RSC org/10.1002/mas.21401. Adv. 2020;10:3092–104. https://doi.org/10.1039/C9RA08985C. 3. Holmes E, Wilson ID, Nicholson JK. Metabolic phenotyping in 20. Li B, He X, Jia W, Li H. Novel applications of metabolomics in health and disease. Cell. 2008;134:714–7. https://doi.org/10.1016/ personalized medicine: a mini-review. Mol Basel Switz. 2017;22: j.cell.2008.08.026. 1173. https://doi.org/10.3390/molecules22071173. 4. da Silva RR, Dorrestein PC, Quinn RA. Illuminating the dark 21. Kohler I, Hankemeier T, van der Graaf PH, Knibbe CAJ, van matter in metabolomics. Proc Natl Acad Sci U S A. 2015;112: Hasselt JGC. Integrating clinical metabolomics-based biomarker 12549–50. https://doi.org/10.1073/pnas.1516878112. discovery and clinical pharmacology to enable precision medi- 5. Psychogios N, Hau DD, Peng J, Guo AC, Mandal R, Bouatra S, cine. Eur J Pharm Sci Off J Eur Fed Pharm Sci. 2017;109S: Sinelnikov I, Krishnamurthy R, Eisner R, Gautam B, Young N, S15–21. https://doi.org/10.1016/j.ejps.2017.05.018. Xia J, Knox C, Dong E, Huang P, Hollander Z, Pedersen TL, 22. Haug K, Salek RM, Conesa P, Hastings J, de Matos P, Rijnbeek Smith SR, Bamforth F, et al. The human serum metabolome. M, Mahendraker T, Williams M, Neumann S, Rocca-Serra P, PLoS ONE. 2011;6:e16957. https://doi.org/10.1371/journal. Maguire E, González-Beltrán A, Sansone S-A, Griffin JL, pone.0016957. Steinbeck C. MetaboLights–an open-access general-purpose re- 6. Patti GJ, Yanes O, Siuzdak G. Innovation: metabolomics: the pository for metabolomics studies and associated meta-data. apogee of the omics trilogy. Nat Rev Mol Cell Biol. 2012;13: Nucleic Acids Res. 2013;41:D781–6. https://doi.org/10.1093/ 263–9. https://doi.org/10.1038/nrm3314. nar/gks1004. 7. Dunn WB, Broadhurst DI, Atherton HJ, Goodacre R, Griffin JL. 23. Haug K, Cochrane K, Nainala VC, Williams M, Chang J, Systems level studies of mammalian metabolomes: the roles of Jayaseelan KV, O’Donovan C. MetaboLights: a resource evolv- mass spectrometry and nuclear magnetic resonance spectroscopy. ing in response to the needs of its scientific community. Nucleic Chem Soc Rev. 2011;40:387–426. https://doi.org/10.1039/ Acids Res. 2020;48:D440–4. https://doi.org/10.1093/nar/ B906712B. gkz1019. Metabolomics for personalized medicine: the input of analytical chemistry from biomarker discovery to... 24. Sud M, Fahy E, Cotter D, Azam K, Vadivelu I, Burant C, Edison Breitkopf SB, Cajka T, Cazenave-Gassiot A, Checa A, et al. A, Fiehn O, Higashi R, Nair KS, Sumner S, Subramaniam S. Harmonizing lipidomics: NIST interlaboratory comparison exer- Metabolomics workbench: an international repository for metabo- cise for lipidomics using SRM 1950–metabolites in frozen human lomics data and metadata, metabolite standards, protocols, tuto- plasma. J Lipid Res. 2017;58:2275–88. https://doi.org/10.1194/ rials and training, and analysis tools. Nucleic Acids Res. 2016;44: jlr.M079012. D463–70. https://doi.org/10.1093/nar/gkv1042. 35. Thompson JW, Adams KJ, Adamski J, Asad Y, Borts D, Bowden 25. Fiehn O, Robertson D, Griffin J, van der Werf M, Nikolau B, JA, Byram G, Dang V, Dunn WB, Fernandez F, Fiehn O, Gaul Morrison N, Sumner LW, Goodacre R, Hardy NW, Taylor C, DA, Hühmer AF, Kalli A, Koal T, Koeniger S, Mandal R, Meier Fostel J, Kristal B, Kaddurah-Daouk R, Mendes P, van Ommen F, Naser FJ, et al. International ring trial of a high resolution B, Lindon JC, Sansone S-A. The metabolomics standards initia- targeted metabolomics and lipidomics platform for serum and tive (MSI). Metabolomics. 2007;3:175–8. https://doi.org/10.1007/ plasma analysis. Anal Chem. 2019;91:14407–16. https://doi.org/ s11306-007-0070-6. 10.1021/acs.analchem.9b02908. 26. Martens L, Chambers M, Sturm M, Kessner D, Levander F, 36. Bearden DW, Sheen DA, Simón-Manso Y, Benner BA, Rocha Shofstahl J, Tang WH, Römpp A, Neumann S, Pizarro AD, WFC, Blonder N, Lippa KA, Beger RD, Schnackenberg LK, Sun Montecchi-Palazzi L, Tasman N, Coleman M, Reisinger F, J, Mehta KY, Cheema AK, Gu H, Marupaka R, Nagana Gowda Souda P, Hermjakob H, Binz P-A, Deutsch EW. mzML–a com- GA, Raftery D. Metabolomics test materials for quality control: a munity standard for mass spectrometry data. Mol Cell Proteomics study of a urine materials suite. Metabolites. 2019;9:270. https:// MCP. 2011;10:R110.000133. https://doi.org/10.1074/mcp.R110. doi.org/10.3390/metabo9110270. 000133. 37. Mandal R, Cano R, Davis CD, Hayashi D, Jackson SA, Jones CM, 27. Schober D, Jacob D, Wilson M, Cruz JA, Marcu A, Grant JR, Lampe JW, Latulippe ME, Lin NJ, Lippa KA, Piotrowski P, Da Moing A, Deborde C, de Figueiredo LF, Haug K, Rocca-Serra P, Silva SM, Swanson KS, Wishart DS. Workshop report: Toward Easton J, Ebbels TMD, Hao J, Ludwig C, Günther UL, Rosato A, the development of a human whole stool reference material for Klein MS, Lewis IA, et al. nmrML: a community supported open metabolomic and metagenomic gut microbiome measurements. data standard for the description, storage, and exchange of NMR Metabolomics Off J Metabolomic Soc. 2020;16:119. https://doi. data. Anal Chem. 2018;90:649–56. https://doi.org/10.1021/acs. org/10.1007/s11306-020-01744-5. analchem.7b02795. 38. Xia J, Sinelnikov IV, Han B, Wishart DS. MetaboAnalyst 3.0– 28. Rocca-Serra P, Salek RM, Arita M, Correa E, Dayalan S, making metabolomics more meaningful. Nucleic Acids Res. Gonzalez-Beltran A, Ebbels T, Goodacre R, Hastings J, Haug 2015;43:W251–7. https://doi.org/10.1093/nar/gkv380. K, Koulman A, Nikolski M, Oresic M, Sansone S-A, Schober 39. Hsu Y-HH, Churchhouse C, Pers TH, Mercader JM, Metspalu A, D, Smith J, Steinbeck C, Viant MR, Neumann S. Data standards Fischer K, Fortney K, Morgen EK, Gonzalez C, Gonzalez ME, can boost metabolomics research, and if there is a will, there is a Esko T, Hirschhorn JN. PAIRUP-MS: Pathway analysis and im- way. Metabolomics. 2016;12:14. https://doi.org/10.1007/s11306- putation to relate unknowns in profiles from mass spectrometry- 015-0879-3. based metabolite data. PLoS Comput Biol. 2019;15:e1006734. 29. Sansone S-A, Rocca-Serra P, Field D, Maguire E, Taylor C, https://doi.org/10.1371/journal.pcbi.1006734. Hofmann O, Fang H, Neumann S, Tong W, Amaral-Zettler L, 40. Pang Z, Zhou G, Chong J, Xia J. Comprehensive meta-analysis of Begley K, Booth T, Bougueleret L, Burns G, Chapman B, Clark COVID-19 global metabolomics datasets. Metabolites. 2021;11. T, Coleman L-A, Copeland J, Das S, et al. Toward interoperable https://doi.org/10.3390/metabo11010044. bioscience data. Nat Genet. 2012;44:121–6. https://doi.org/10. 41. Ioannidis JPA, Bossuyt PMM. Waste, leaks, and failures in the 1038/ng.1054. biomarker pipeline. Clin Chem. 2017;63:963–72. https://doi.org/ 30. Rampler E, Abiead YE, Schoeny H, Rusz M, Hildebrand F, Fitz 10.1373/clinchem.2016.254649. V, Koellensperger G. Recurrent topics in mass spectrometry- 42. Poste G. Bring on the biomarkers. Nature. 2011;469:156–7. based metabolomics and lipidomics-standardization, coverage, https://doi.org/10.1038/469156a. and throughput. Anal Chem. 2021;93:519–45. https://doi.org/10. 43. Ramautar R, Berger R, van der Greef J, Hankemeier T. Human 1021/acs.analchem.0c04698. metabolomics: strategies to understand biology. Curr Opin Chem 31. Phinney KW, Ballihaut G, Bedner M, Benford BS, Camara JE, Biol. 2013;17:841–6. https://doi.org/10.1016/j.cbpa.2013.06.015. Christopher SJ, Davis WC, Dodder NG, Eppe G, Lang BE, Long 44. Kennedy AD, Wittmann BM, Evans AM, Miller LAD, Toal DR, SE, Lowenthal MS, McGaw EA, Murphy KE, Nelson BC, Lonergan S, Elsea SH, Pappan KL. Metabolomics in the clinic: a Prendergast JL, Reiner JL, Rimmer CA, Sander LC, et al. review of the shared and unique features of untargeted metabolo- Development of a standard reference material for metabolomics mics for clinical research and clinical testing. J Mass Spectrom research. Anal Chem. 2013;85:11732–8. https://doi.org/10.1021/ JMS. 2018;53:1143–54. https://doi.org/10.1002/jms.4292. ac402689t. 45. Collino S, Martin F-PJ, Rezzi S. Clinical metabolomics paves the 32. Siskos AP, Jain P, Römisch-Margl W, Bennett M, Achaintre D, way towards future healthcare strategies. Br J Clin Pharmacol. Asad Y, Marney L, Richardson L, Koulman A, Griffin JL, 2013;75:619–29. https://doi.org/10.1111/j.1365-2125.2012. Raynaud F, Scalbert A, Adamski J, Prehn C, Keun HC. 04216.x. Interlaboratory reproducibility of a targeted metabolomics plat- 46. Beger RD, Dunn W, Schmidt MA, Gross SS, Kirwan JA, form for analysis of human serum and plasma. Anal Chem. Cascante M, Brennan L, Wishart DS, Oresic M, Hankemeier T, 2017;89:656–65. https://doi.org/10.1021/acs.analchem.6b02930. Broadhurst DI, Lane AN, Suhre K, Kastenmüller G, Sumner SJ, 33. Telu KH, Yan X, Wallace WE, Stein SE, Simón-Manso Y. Thiele I, Fiehn O, Kaddurah-Daouk R, for “Precision Medicine Analysis of human plasma metabolites across different liquid and Pharmacometabolomics Task Group”-Metabolomics Society chromatography/mass spectrometry platforms: cross-platform Initiative. Metabolomics enables precision medicine: “A White transferable chemical signatures. Rapid Commun Mass Paper, Community Perspective”. Metabolomics Off J Spectrom RCM. 2016;30:581–93. https://doi.org/10.1002/rcm. Metabolomic Soc. 2016;12:149. https://doi.org/10.1007/s11306- 7475. 016-1094-6. 34. Bowden JA, Heckert A, Ulmer CZ, Jones CM, Koelmel JP, 47. Saito K, Maekawa K, Pappan KL, Urata M, Ishikawa M, Kumagai Abdullah L, Ahonen L, Alnouti Y, Armando AM, Asara JM, Y, Saito Y. Differences in metabolite profiles between blood ma- Bamba T, Barr JR, Bergquist J, Borchers CH, Brandsma J, trices, ages, and sexes among Caucasian individuals and their Castelli F.A. et al. inter-individual variations. Metabolomics. 2014;10:402–13. Nicholls AW, Wilson ID, Kell DB, Goodacre R, Human Serum https://doi.org/10.1007/s11306-013-0591-0. Metabolome (HUSERMET) Consortium. Procedures for large- 48. Wedge DC, Allwood JW, Dunn W, Vaughan AA, Simpson K, scale metabolic profiling of serum and plasma using gas chroma- Brown M, Priest L, Blackhall FH, Whetton AD, Dive C, tography and liquid chromatography coupled to mass spectrome- Goodacre R. Is serum or plasma more appropriate for intersubject try. Nat Protoc. 2011;6:1060–83. https://doi.org/10.1038/nprot. comparisons in metabolomic studies? An assessment in patients 2011.335. with small-cell lung cancer. Anal Chem. 2011;83:6689–97. 62. González-Riano C, Dudzik D, Garcia A, Gil-de-la-Fuente A, https://doi.org/10.1021/ac2012224. Gradillas A, Godzien J, López-Gonzálvez Á, Rey-Stolle F, Rojo 49. Fang M, Ivanisevic J, Benton HP, Johnson CH, Patti GJ, Hoang D, Ruperez FJ, Saiz J, Barbas C. Recent developments along the LT, Uritboonthai W, Kurczy ME, Siuzdak G. Thermal degrada- analytical process for metabolomics workflows. Anal Chem. tion of small molecules: a global metabolomic investigation. Anal 2020;92:203–26. https://doi.org/10.1021/acs.analchem.9b04553. Chem. 2015;87:10935–41. https://doi.org/10.1021/acs.analchem. 63. Boudah S, Olivier M-F, Aros-Calt S, Oliveira L, Fenaille F, Tabet 5b03003. J-C, Junot C. Annotation of the human serum metabolome by 50. Roux A, Thévenot EA, Seguin F, Olivier M-F, Junot C. Impact of coupling three liquid chromatography methods to high- collection conditions on the metabolite content of human urine resolution mass spectrometry. J Chromatogr B Analyt Technol samples as analyzed by liquid chromatography coupled to mass Biomed Life Sci. 2014;966:34–47. https://doi.org/10.1016/j. spectrometry and nuclear magnetic resonance spectroscopy. jchromb.2014.04.025. Metabolomics. 2015;11:1095–105. https://doi.org/10.1007/ 64. Karu N, Deng L, Slae M, Guo AC, Sajed T, Huynh H, Wine E, s11306-014-0764-5. Wishart DS. A review on human fecal metabolomics: methods, 51. Lu W, Su X, Klein MS, Lewis IA, Fiehn O, Rabinowitz JD. applications and the human fecal metabolome database. Anal Metabolite measurement: pitfalls to avoid and practices to follow. Chim Acta. 2018;1030:1–24. https://doi.org/10.1016/j.aca.2018. Annu Rev Biochem. 2017;86:277–304. https://doi.org/10.1146/ 05.031. annurev-biochem-061516-044952. 65. Tabone M, Bressa C, García-Merino JA, Moreno-Pérez D, Van 52. Kamlage B, Maldonado SG, Bethan B, Peter E, Schmitz O, EC, Castelli FA, Fenaille F, Larrosa M. The effect of acute Liebenberg V, Schatz P. Quality markers addressing preanalytical moderate-intensity exercise on the serum and fecal metabolomes variations of blood and plasma processing identified by broad and and the gut microbiota of cross-country endurance athletes. Sci targeted metabolite profiling. Clin Chem. 2014;60:399–412. Rep. 2021;11:3558. https://doi.org/10.1038/s41598-021-82947-1. https://doi.org/10.1373/clinchem.2013.211979. 66. Mervant L, Tremblay-Franco M, Jamin EL, Kesse-Guyot E, 53. Rebholz CM, Rhee EP. Application of metabolomics to renal and Galan P, Martin J-F, Guéraud F, Debrauwer L. Osmolality- cardiometabolic diseases. Methods Mol Biol Clifton NJ. based normalization enhances statistical discrimination of 2020;2104:401–17. https://doi.org/10.1007/978-1-0716-0239-3_ untargeted metabolomic urine analysis: results from a compara- 20. tive study. Metabolomics Off J Metabolomic Soc. 2021;17:2. 54. Blaise BJ, Correia G, Tin A, Young JH, Vergnaud A-C, Lewis M, https://doi.org/10.1007/s11306-020-01758-z. Pearce JTM, Elliott P, Nicholson JK, Holmes E, Ebbels TMD. 67. Thévenot EA, Roux A, Xu Y, Ezan E, Junot C. Analysis of the Power analysis and sample size determination in metabolic phe- human adult urinary metabolome variations with age, body mass notyping. Anal Chem. 2016;88:5179–88. https://doi.org/10.1021/ index, and gender by implementing a comprehensive workflow acs.analchem.6b00188. for univariate and OPLS statistical analyses. J Proteome Res. 55. Nyamundanda G, Gormley IC, Fan Y, Gallagher WM, Brennan L. 2015;14:3322–35. https://doi.org/10.1021/acs.jproteome. MetSizeR: selecting the optimal sample size for metabolomic 5b00354. studies using an analysis based approach. BMC Bioinformatics. 68. Edmands WMB, Ferrari P, Scalbert A. Normalization to specific 2013;14:338. https://doi.org/10.1186/1471-2105-14-338. gravity prior to analysis improves information recovery from high 56. Billoir E, Navratil V, Blaise BJ. Sample size calculation in meta- resolution mass spectrometry metabolomic profiles of human bolic phenotyping studies. Brief Bioinform. 2015;16:813–9. urine. Anal Chem. 2014;86:10925–31. https://doi.org/10.1021/ https://doi.org/10.1093/bib/bbu052. ac503190m. 57. Kirwan JA, Brennan L, Broadhurst D, Fiehn O, Cascante M, 69. Chen Y, Shen G, Zhang R, He J, Zhang Y, Xu J, Yang W, Chen Dunn WB, Schmidt MA, Velagapudi V. Preanalytical processing X, Song Y, Abliz Z. Combination of injection volume calibration and biobanking procedures of biological samples for metabolo- by creatinine and MS signals’ normalization to overcome urine mics research: a white paper, community perspective (for variability in LC-MS-based metabolomics studies. Anal Chem. “Precision Medicine and Pharmacometabolomics Task Group”- 2013;85:7659–65. https://doi.org/10.1021/ac401400b. The Metabolomics Society Initiative). Clin Chem. 2018;64: 70. Warrack BM, Hnatyshyn S, Ott K-H, Reily MD, Sanders M, 1158–82. https://doi.org/10.1373/clinchem.2018.287045. Zhang H, Drexler DM. Normalization strategies for metabonomic 58. Long NP, Nghi TD, Kang YP, Anh NH, Kim HM, Park SK, analysis of urine samples. J Chromatogr B Analyt Technol Kwon SW. Toward a standardized strategy of clinical metabolo- Biomed Life Sci. 2009;877:547–52. https://doi.org/10.1016/j. mics for the advancement of precision medicine. Metabolites. jchromb.2009.01.007. 2020;10:51. https://doi.org/10.3390/metabo10020051. 71. Jacob CC, Dervilly-Pinel G, Biancotto G, Le Bizec B. Evaluation 59. Dudzik D, Barbas-Bernardos C, García A, Barbas C. Quality as- of specific gravity as normalization strategy for cattle urinary me- surance procedures for mass spectrometry untargeted metabolo- tabolome analysis. Metabolomics. 2014;10:627–37. https://doi. mics. A review. J Pharm Biomed Anal. 2018;147:149–73. https:// org/10.1007/s11306-013-0604-z. doi.org/10.1016/j.jpba.2017.07.044. 72. Meister I, Zhang P, Sinha A, Sköld CM, Wheelock ÅM, Izumi T, 60. Pepe MS, Li CI, Feng Z. Improving the quality of biomarker Chaleckis R, Wheelock CE. High-precision automated workflow discovery research: the right samples and enough of them. for urinary untargeted metabolomic epidemiology. Anal Chem. Cancer Epidemiol Biomark Prev Publ Am Assoc Cancer Res 2021;93:5248–58. https://doi.org/10.1021/acs.analchem. Cosponsored Am Soc Prev Oncol. 2015;24:944–50. https://doi. 1c00203. org/10.1158/1055-9965.EPI-14-1227. 73. Aros-Calt S, Castelli FA, Lamourette P, Gervasi G, Junot C, 61. Dunn WB, Broadhurst D, Begley P, Zelena E, Francis-McIntyre Muller BH, Fenaille F. Metabolomic investigation of S, Anderson N, Brown M, Knowles JD, Halsall A, Haselden JN, Staphylococcus aureus antibiotic susceptibility by liquid Metabolomics for personalized medicine: the input of analytical chemistry from biomarker discovery to... chromatography coupled to high-resolution mass spectrometry. 86. Wang M, Carver JJ, Phelan VV, Sanchez LM, Garg N, Peng Y, Methods Mol Biol Clifton NJ. 2019;1871:279–93. https://doi. Nguyen DD, Watrous J, Kapono CA, Luzzatto-Knaan T, Porto C, org/10.1007/978-1-4939-8814-3_18. Bouslimani A, Melnik AV, Meehan MJ, Liu W-T, Crüsemann M, 74. Moyne O, Castelli F, Bicout DJ, Boccard J, Camara B, Cournoyer Boudreau PD, Esquenazi E, Sandoval-Calderón M, et al. Sharing B, Faudry E, Terrier S, Hannani D, Huot-Marchand S, Léger C, and community curation of mass spectrometry data with Global Maurin M, Ngo T-D, Plazy C, Quinn RA, Attree I, Fenaille F, Natural Products Social Molecular Networking. Nat Biotechnol. Toussaint B, Le Gouëllec A. Metabotypes of Pseudomonas 2016;34:828–37. https://doi.org/10.1038/nbt.3597. aeruginosa correlate with antibiotic resistance, virulence and clin- 87. Fahy E, Sud M, Cotter D, Subramaniam S. LIPID MAPS online ical outcome in cystic fibrosis chronic infections. Metabolites. tools for lipid research. Nucleic Acids Res. 2007;35:W606–12. 2021;11:63. https://doi.org/10.3390/metabo11020063. https://doi.org/10.1093/nar/gkm324. 75. Contrepois K, Jiang L, Snyder M. Optimized analytical proce- 88. Smith CA, O’Maille G, Want EJ, Qin C, Trauger SA, Brandon dures for the untargeted metabolomic profiling of human urine TR, Custodio DE, Abagyan R, Siuzdak G. METLIN: a metabolite and plasma by combining hydrophilic interaction (HILIC) and mass spectral database. Ther Drug Monit. 2005;27:747–51. reverse-phase liquid chromatography (RPLC)-mass spectrometry. 89. Vinaixa M, Schymanski EL, Neumann S, Navarro M, Salek RM, Mol Cell Proteomics MCP. 2015;14:1684–95. https://doi.org/10. Yanes O. Mass spectral databases for LC/MS- and GC/MS-based 1074/mcp.M114.046508. metabolomics: state of the field and future prospects. TrAC Trends 76. Clish CB. Metabolomics: an emerging but powerful tool for pre- Anal Chem. 2016;78:23–35. https://doi.org/10.1016/j.trac.2015. cision medicine. Cold Spring Harb Mol Case Stud. 2015;1: 09.005. a000588. https://doi.org/10.1101/mcs.a000588. 90. Frainay C, Schymanski EL, Neumann S, Merlet B, Salek RM, 77. Southam AD, Haglington LD, Najdekr L, Jankevics A, Weber Jourdan F, Yanes O. Mind the gap: mapping mass spectral data- RJM, Dunn WB. Assessment of human plasma and urine sample bases in genome-scale metabolic networks reveals poorly covered preparation for reproducible and high-throughput UHPLC-MS areas. Metabolites. 2018;8:51. https://doi.org/10.3390/ clinical metabolic phenotyping. The Analyst. 2020;145:6511– metabo8030051. 23. https://doi.org/10.1039/d0an01319f. 91. Blaženović I, Kind T, Sa MR, Ji J, Vaniya A, Wancewicz B, 78. Blaženović I, Kind T, Ji J, Fiehn O. Software tools and approaches Roberts BS, Torbašinović H, Lee T, Mehta SS, Showalter MR, for compound identification of LC-MS/MS data in metabolomics. Song H, Kwok J, Jahn D, Kim J, Fiehn O. Structure annotation of Metabolites. 2018;8:31. https://doi.org/10.3390/metabo8020031. all mass spectra in untargeted metabolomics. Anal Chem. 2019;91:2155–62. https://doi.org/10.1021/acs.analchem. 79. Moreau R, Clària J, Aguilar F, Fenaille F, Lozano JJ, Junot C, 8b04698. Colsch B, Caraceni P, Trebicka J, Pavesi M, Alessandria C, 92. Oberacher H, Sasse M, Antignac J-P, Guitton Y, Debrauwer L, Nevens F, Saliba F, Welzel TM, Albillos A, Gustot T, Jamin EL, Schulze T, Krauss M, Covaci A, Caballero-Casero N, Fernández J, Moreno C, Baldassarre M, et al. Blood metabolomics Rousseau K, Damont A, Fenaille F, Lamoree M, Schymanski EL. uncovers inflammation-associated mitochondrial dysfunction as a A European proposal for quality control and quality assurance of potential mechanism underlying ACLF. J Hepatol. 2020;72:688– tandem mass spectral libraries. Environ Sci Eur. 2020;32. https:// 701. https://doi.org/10.1016/j.jhep.2019.11.009. doi.org/10.1186/s12302-020-00314-9. 80. Zaccherini G, Aguilar F, Caraceni P, Clària J, Lozano JJ, Fenaille 93. Wilkinson MD, Dumontier M, Sansone S-A, Bonino da Silva F, Castelli F, Junot C, Curto A, Formentin C, Weiss E, Bernardi Santos LO, Prieto M, Batista D, Mcquilton P, Kuhn T, Rocca- M, Jalan R, Angeli P, Moreau R, Arroyo V. Assessing the role of Serra P, Crosas M, Schultes E. Evaluating FAIR maturity through amino acids in systemic inflammation and organ failure in patients a scalable, automated, community-governed framework. Sci Data. with ACLF. J Hepatol. 2021;74:1117–31. https://doi.org/10.1016/ 2019;6:174. https://doi.org/10.1038/s41597-019-0184-5. j.jhep.2020.11.035. 94. Damont A, Olivier M-F, Warnet A, Lyan B, Pujos-Guillot E, 81. Kind T, Fiehn O. Metabolomic database annotations via query of Jamin EL, Debrauwer L, Bernillon S, Junot C, Tabet J-C, elemental compositions: mass accuracy is insufficient even at less Fenaille F. Proposal for a chemically consistent way to annotate than 1 ppm. BMC Bioinformatics. 2006;7:234. https://doi.org/10. ions arising from the analysis of reference compounds under ESI 1186/1471-2105-7-234. conditions: a prerequisite to proper mass spectral database consti- 82. Nash WJ, Dunn WB. From mass to metabolite in human tution in metabolomics. J Mass Spectrom JMS. 2019;54:567–82. untargeted metabolomics: recent advances in annotation of me- https://doi.org/10.1002/jms.4372. tabolites applying liquid chromatography-mass spectrometry da- 95. Peisl BYL, Schymanski EL, Wilmes P. Dark matter in host- ta. TrAC Trends Anal Chem. 2019;120:115324. https://doi.org/ microbiome metabolomics: tackling the unknowns-a review. 10.1016/j.trac.2018.11.022. Anal Chim Acta. 2018;1037:13–27. https://doi.org/10.1016/j. 83. Barbier Saint Hilaire P, Hohenester UM, Colsch B, Tabet J-C, aca.2017.12.034. Junot C, Fenaille F. Evaluation of the high-field Orbitrap fusion 96. Simón-Manso Y, Marupaka R, Yan X, Liang Y, Telu KH, for compound annotation in metabolomics. Anal Chem. 2018;90: Mirokhin Y, Stein SE. Mass spectrometry fingerprints of small- 3030–5. https://doi.org/10.1021/acs.analchem.7b05372. molecule metabolites in biofluids: building a spectral library of 84. Horai H, Arita M, Kanaya S, Nihei Y, Ikeda T, Suwa K, Ojima Y, recurrent spectra for urine analysis. Anal Chem. 2019;91:12021– Tanaka K, Tanaka S, Aoshima K, Oda Y, Kakazu Y, Kusano M, 9. https://doi.org/10.1021/acs.analchem.9b02977. Tohge T, Matsuda F, Sawada Y, Hirai MY, Nakanishi H, Ikeda K, 97. Li S, Park Y, Duraisingham S, Strobel FH, Khan N, Soltow QA, et al. MassBank: a public repository for sharing mass spectral data Jones DP, Pulendran B. Predicting network activity from high for life sciences. J Mass Spectrom JMS. 2010;45:703–14. https:// throughput metabolomics. PLoS Comput Biol. 2013;9: doi.org/10.1002/jms.1777. e1003123. https://doi.org/10.1371/journal.pcbi.1003123. 85. Wishart DS, Tzur D, Knox C, Eisner R, Guo AC, Young N, 98. Broadhurst D, Goodacre R, Reinke SN, Kuligowski J, Wilson ID, Cheng D, Jewell K, Arndt D, Sawhney S, Fung C, Nikolai L, Lewis MR, Dunn WB. Guidelines and considerations for the use Lewis M, Coutouly M-A, Forsythe I, Tang P, Shrivastava S, of system suitability and quality control samples in mass spec- Jeroncic K, Stothard P, et al. HMDB: the human metabolome trometry assays applied in untargeted clinical metabolomic stud- database. Nucleic Acids Res. 2007;35:D521–6. https://doi.org/ ies. Metabolomics. 2018;14:72. https://doi.org/10.1007/s11306- 10.1093/nar/gkl923. 018-1367-3. Castelli F.A. et al. 99. Evans AM, O’Donovan C, Playdon M, Beecher C, Beger RD, Rocca-Serra P, Salek RM, Walk T, Weber RJM. Use cases, best Bowden JA, Broadhurst D, Clish CB, Dasari S, Dunn WB, practice and reporting standards for metabolomics in regulatory Griffin JL, Hartung T, Hsu P-C, Huan T, Jans J, Jones CM, toxicology. Nat Commun. 2019;10:3041. https://doi.org/10.1038/ Kachman M, Kleensang A, Lewis MR, et al. Dissemination and s41467-019-10900-y. analysis of the quality assurance (QA) and quality control (QC) 112. Burla B, Arita M, Arita M, Bendt AK, Cazenave-Gassiot A, practices of LC-MS based untargeted metabolomics practitioners. Dennis EA, Ekroos K, Han X, Ikeda K, Liebisch G, Lin MK, Metabolomics Off J Metabolomic Soc. 2020;16:113. https://doi. Loh TP, Meikle PJ, Orešič M, Quehenberger O, Shevchenko A, org/10.1007/s11306-020-01728-5. Torta F, Wakelam MJO, Wheelock CE, Wenk MR. MS-based 100. Sangster T, Major H, Plumb R, Wilson AJ, Wilson ID. A prag- lipidomics of human blood plasma: a community-initiated posi- matic and readily implemented quality control strategy for HPLC- tion paper to develop accepted guidelines. J Lipid Res. 2018;59: MS and GC-MS-based metabonomic analysis. The Analyst. 2001–17. https://doi.org/10.1194/jlr.S087163. 2006;131:1075–8. https://doi.org/10.1039/b604498k. 113. Beger RD, Dunn WB, Bandukwala A, Bethan B, Broadhurst D, 101. Godzien J, Kalaska B, Adamska-Patruno E, Siroka J, Ciborowski Clish CB, Dasari S, Derr L, Evans A, Fischer S, Flynn T, Hartung M, Kretowski A, Barbas C. Oxidized T, Herrington D, Higashi R, Hsu P-C, Jones C, Kachman M, glycerophosphatidylcholines in diabetes through non-targeted Karuso H, Kruppa G, et al. Towards quality assurance and quality metabolomics: their annotation and biological meaning. J control in untargeted metabolomics studies. Metabolomics Off J Chromatogr B Analyt Technol Biomed Life Sci. 2019;1120:62– Metabolomic Soc. 2019;15:4. https://doi.org/10.1007/s11306- 70. https://doi.org/10.1016/j.jchromb.2019.04.053. 018-1460-7. 102. Rafiei A, Sleno L. Comparison of peak-picking workflows for 114. Laparre J, Kaabia Z, Mooney M, Buckley T, Sherry M, Le Bizec untargeted liquid chromatography/high-resolution mass spectrom- B, Dervilly-Pinel G. Impact of storage conditions on the urinary etry metabolomics data analysis. Rapid Commun Mass Spectrom metabolomics fingerprint. Anal Chim Acta. 2017;951:99–107. RCM. 2015;29:119–27. https://doi.org/10.1002/rcm.7094. https://doi.org/10.1016/j.aca.2016.11.055. 103. Giacomoni F, Le Corguillé G, Monsoor M, Landi M, Pericard P, 115. Palmer EA, Cooper HJ, Dunn WB. Investigation of the 12-month Pétéra M, Duperier C, Tremblay-Franco M, Martin J-F, Jacob D, stability of dried blood and urine spots applying untargeted Goulitquer S, Thévenot EA, Caron C. Workflow4Metabolomics: UHPLC-MS metabolomic assays. Anal Chem. 2019;91:14306– a collaborative research infrastructure for computational metabo- 13. https://doi.org/10.1021/acs.analchem.9b02577. lomics. Bioinforma Oxf Engl. 2015;31:1493–5. https://doi.org/10. 116. Haid M, Muschet C, Wahl S, Römisch-Margl W, Prehn C, Möller 1093/bioinformatics/btu813. G, Adamski J. Long-term stability of human plasma metabolites 104. Chong J, Soufan O, Li C, Caraus I, Li S, Bourque G, Wishart DS, during storage at -80 °C. J Proteome Res. 2018;17:203–11. https:// Xia J. MetaboAnalyst 4.0: towards more transparent and integra- doi.org/10.1021/acs.jproteome.7b00518. tive metabolomics analysis. Nucleic Acids Res. 2018;46:W486– 117. Wagner-Golbs A, Neuber S, Kamlage B, Christiansen N, Bethan 94. https://doi.org/10.1093/nar/gky310. B, Rennefahrt U, Schatz P, Lind L. Effects of long-term storage at 105. Sumner LW, Amberg A, Barrett D, Beale MH, Beger R, Daykin -80 °C on the human plasma metabolome. Metabolites. 2019;9: CA, Fan TW-M, Fiehn O, Goodacre R, Griffin JL, Hankemeier T, 884–96. https://doi.org/10.3390/metabo9050099. Hardy N, Harnly J, Higashi R, Kopka J, Lane AN, Lindon JC, Marriott P, Nicholls AW, et al. Proposed minimum reporting stan- 118. Dunn WB, Lin W, Broadhurst D, Begley P, Brown M, Zelena E, dards for chemical analysis Chemical Analysis Working Group Vaughan AA, Halsall A, Harding N, Knowles JD, Francis- (CAWG) Metabolom ics Sta ndards Initiative (MSI). McIntyre S, Tseng A, Ellis DI, O’Hagan S, Aarons G, Benjamin Metabolomics Off J Metabolomic Soc. 2007;3:211–21. https:// B, Chew-Graham S, Moseley C, Potter P, et al. Molecular pheno- doi.org/10.1007/s11306-007-0082-2. typing of a UK population: defining the human serum metabo- 106. Creek DJ, Dunn WB, Fiehn O, Griffin JL, Hall RD, Lei Z, Mistrik lome. Metabolomics Off J Metabolomic Soc. 2015;11:9–26. R, Neumann S, Schymanski EL, Sumner LW, Trengove R, https://doi.org/10.1007/s11306-014-0707-1. Wolfender J-L. Metabolite identification: are you sure? And how 119. Sindelar M, Stancliffe E, Schwaiger-Haber M, Anbukumar DS, do your peers gauge your confidence? Metabolomics. 2014;10: Albrecht RA, Liu W-C, Adkins-Travis K, Garcia-Sastre A, 350–3. https://doi.org/10.1007/s11306-014-0656-8. Shriver LP, Patti GJ. Longitudinal metabolomics of human plasma 107. Schymanski EL, Jeon J, Gulde R, Fenner K, Ruff M, Singer HP, reveals robust prognostic markers of COVID-19 disease severity. Hollender J. Identifying small molecules via high resolution mass MedRxiv Prepr Serv Health Sci. 2021. https://doi.org/10.1101/ spectrometry: communicating confidence. Environ Sci Technol. 2021.02.05.21251173. 2014;48:2097–8. https://doi.org/10.1021/es5002105. 120. Fernández-Albert F, Llorach R, Garcia-Aloy M, Ziyatdinov A, 108. Sumner LW, Lei Z, Nikolau BJ, Saito K, Roessner U, Trengove R. Andres-Lacueva C, Perera A. Intensity drift removal in LC/MS Proposed quantitative and alphanumeric metabolite identification metabolomics by common variance compensation. Bioinforma metrics. Metabolomics. 2014;10:1047–9. https://doi.org/10.1007/ Oxf Engl. 2014;30:2899–905. https://doi.org/10.1093/ s11306-014-0739-6. bioinformatics/btu423. 109. Lewis MR, Pearce JTM, Spagou K, Green M, Dona AC, Yuen 121. Brunius C, Shi L, Landberg R. Large-scale untargeted LC-MS AHY, David M, Berry DJ, Chappell K, Horneffer-van der Sluis V, metabolomics data correction using between-batch feature align- Shaw R, Lovestone S, Elliott P, Shockcor J, Lindon JC, Cloarec ment and cluster-based within-batch signal intensity drift correc- O, Takats Z, Holmes E, Nicholson JK. Development and applica- tion. Metabolomics Off J Metabolomic Soc. 2016;12:173. https:// tion of ultra-performance liquid chromatography-TOF MS for pre- doi.org/10.1007/s11306-016-1124-4. cision large scale urinary metabolic phenotyping. Anal Chem. 122. Webb-Robertson B-JM, Stratton KG, Kyle JE, Kim Y-M, Bramer 2016;88:9004–13. https://doi.org/10.1021/acs.analchem. LM, Waters KM, Koeller DM, Metz TO. Statistically driven me- 6b01481. tabolite and lipid profiling of patients from the undiagnosed dis- 110. Considine EC. The search for clinically useful biomarkers of com- eases network. Anal Chem. 2020;92:1796–803. https://doi.org/10. plex disease: a data analysis perspective. Metabolites. 2019;9:126. 1021/acs.analchem.9b03522. https://doi.org/10.3390/metabo9070126. 123. Watrous JD, Henglin M, Claggett B, Lehmann KA, Larson MG, 111. Viant MR, Ebbels TMD, Beger RD, Ekman DR, Epps DJT, Kamp Cheng S, Jain M. Visualization, quantification, and alignment of H, Leonards PEG, Loizou GD, MacRae JI, van Ravenzwaay B, spectral drift in population scale untargeted metabolomics data. Metabolomics for personalized medicine: the input of analytical chemistry from biomarker discovery to... Anal Chem. 2017;89:1399–404. https://doi.org/10.1021/acs. 136. Cho K, Schwaiger-Haber M, Naser FJ, Stancliffe E, Sindelar M, analchem.6b04337. Patti GJ. Targeting unique biological signals on the fly to improve 124. Liu Q, Walker D, Uppal K, Liu Z, Ma C, Tran V, Li S, Jones DP, MS/MS coverage and identification efficiency in metabolomics. Yu T. Addressing the batch effect issue for LC/MS metabolomics Anal Chim Acta. 2021;1149:338210. https://doi.org/10.1016/j. data in data preprocessing. Sci Rep. 2020;10:13856. https://doi. aca.2021.338210. org/10.1038/s41598-020-70850-0. 137. Zha H, Cai Y, Yin Y, Wang Z, Li K, Zhu Z-J. SWATHtoMRM: 125. Li S, Sullivan NL, Rouphael N, Yu T, Banton S, Maddur MS, development of high-coverage targeted metabolomics method McCausland M, Chiu C, Canniff J, Dubey S, Liu K, Tran V, using SWATH technology for biomarker discovery. Anal Chem. Hagan T, Duraisingham S, Wieland A, Mehta AK, Whitaker JA, 2018;90:4062–70. https://doi.org/10.1021/acs.analchem. Subramaniam S, Jones DP, et al. Metabolic phenotypes of re- 7b05318. sponse to vaccination in humans. Cell. 2017;169:862–877.e17. 138. Rinaudo P, Boudah S, Junot C, Thévenot EA. biosigner: a new https://doi.org/10.1016/j.cell.2017.04.026. method for the discovery of significant molecular signatures from 126. Martin J-C, Maillot M, Mazerolles G, Verdu A, Lyan B, Migné C, omics data. Front Mol Biosci 3:26. 2016. https://doi.org/10.3389/ Defoort C, Canlet C, Junot C, Guillou C, Manach C, Jabob D, fmolb.2016.00026. Bouveresse DJ-R, Paris E, Pujos-Guillot E, Jourdan F, Giacomoni 139. Pirro V, Alfaro CM, Jarmusch AK, Hattab EM, Cohen-Gadol AA, F, Courant F, Favé G, et al. Can we trust untargeted metabolo- Cooks RG. Intraoperative assessment of tumor margins during mics? Results of the metabo-ring initiative, a large-scale, multi- glioma resection by desorption electrospray ionization-mass spec- instrument inter-laboratory study. Metabolomics. 2015;11:807– trometry. Proc Natl Acad Sci U S A. 2017;114:6700–5. https://doi. 21. https://doi.org/10.1007/s11306-014-0740-0. org/10.1073/pnas.1706459114. 127. Izumi Y, Matsuda F, Hirayama A, Ikeda K, Kita Y, Horie K, 140. Brown HM, Alfaro CM, Pirro V, Dey M, Hattab EM, Cohen- Saigusa D, Saito K, Sawada Y, Nakanishi H, Okahashi N, Gadol AA, Cooks RG. Intraoperative mass spectrometry platform Takahashi M, Nakao M, Hata K, Hoshi Y, Morihara M, Tanabe for IDH mutation status prediction, glioma diagnosis, and estima- K, Bamba T, Oda Y. Inter-laboratory comparison of metabolite tion of tumor cell infiltration. J Appl Lab Med. 2021. https://doi. measurements for metabolomics data integration. Metabolites. org/10.1093/jalm/jfaa233. 2019;9:257. https://doi.org/10.3390/metabo9110257. 141. Iwata T, Yamasaki K. Enzymatic determination and thin-layer 128. Yu B, Zanetti KA, Temprosa M, Albanes D, Appel N, Barrera CB, chromatography of bile acids in blood. J Biochem (Tokyo). Ben-Shlomo Y, Boerwinkle E, Casas JP, Clish C, Dale C, 1964;56:424–31. https://doi.org/10.1093/oxfordjournals.jbchem. Dehghan A, Derkach A, Eliassen AH, Elliott P, Fahy E, Gieger a128013. C, Gunter MJ, Harada S, et al. The Consortium of Metabolomics 142. Triebig G, Schaller KH. A simple and reliable enzymatic assay for Studies (COMETS): metabolomics in 47 prospective cohort stud- the determination of formic acid in urine. Clin Chim Acta Int J ies. Am J Epidemiol. 2019;188:991–1012. https://doi.org/10. Clin Chem. 1980;108:355–60. https://doi.org/10.1016/0009- 1093/aje/kwz028. 8981(80)90341-1. 129. Roberts LD, Souza AL, Gerszten RE, Clish CB. Targeted meta- 143. Rolton HA, McConnell KN, Modi KS, Macdougall AI. A simple, bolomics. Curr Protoc Mol Biol Chapter. 2012;30:Unit 30.2.1-24. rapid assay for plasma oxalate in uraemic patients using oxalate https://doi.org/10.1002/0471142727.mb3002s98. oxidase, which is free from vitamin C interference. Clin Chim 130. Clària J, Moreau R, Fenaille F, Amorós A, Junot C, Gronbaek H, Acta Int J Clin Chem. 1989;182:247–54. https://doi.org/10.1016/ Coenraad MJ, Pruvost A, Ghettas A, Chu-Van E, López-Vicario 0009-8981(89)90102-2. C, Oettl K, Caraceni P, Alessandria C, Trebicka J, Pavesi M, 144. Crook MA, Kargbo S, Lumb P. Measurement of urine total sialic Deulofeu C, Albillos A, Gustot T, et al. Orchestration of acid: comparison of an automated ultraviolet enzymatic method tryptophan-kynurenine pathway, acute decompensation, and with a colorimetric assay. Br J Biomed Sci. 2002;59:20–3. https:// acute-on-chronic liver failure in cirrhosis. Hepatology. 2019;69: doi.org/10.1080/09674845.2002.11783629. 1686–701. https://doi.org/10.1002/hep.30363. 145. Žížalová K, Vecka M, Vítek L, Leníček M. Enzymatic methods 131. Raetz M, Bonner R, Hopfgartner G. SWATH-MS for metabolo- may underestimate the total serum bile acid concentration. PloS mics and lipidomics: critical aspects of qualitative and quantitative One. 2020;15:e0236372. https://doi.org/10.1371/journal.pone. analysis. Metabolomics Off J Metabolomic Soc. 2020;16:71. 0236372. https://doi.org/10.1007/s11306-020-01692-0. 146. Darwish IA. Immunoassay methods and their applications in phar- 132. Fenaille F, Barbier Saint-Hilaire P, Rousseau K, Junot C. Data maceutical analysis: basic methodology and recent advances. Int J acquisition workflows in liquid chromatography coupled to high Biomed Sci IJBS. 2006;2:217–35. resolution mass spectrometry-based metabolomics: where do we 147. Pratt JJ. Steroid immunoassay in clinical chemistry. Clin Chem. stand? J Chromatogr A. 2017;1526:1–12. https://doi.org/10.1016/ 1978;24:1869–90. j.chroma.2017.10.043. 148. Wudy SA, Schuler G, Sánchez-Guijo A, Hartmann MF. The art of 133. Barbier Saint Hilaire P, Rousseau K, Seyer A, Dechaumet S, measuring steroids: principles and practice of current hormonal Damont A, Junot C, Fenaille F. Comparative evaluation of data steroid analysis. J Steroid Biochem Mol Biol. 2018;179:88–103. dependent and data independent acquisition workflows imple- https://doi.org/10.1016/j.jsbmb.2017.09.003. mented on an Orbitrap fusion for untargeted metabolomics. 149. Holick MF. Vitamin D status: measurement, interpretation, and M e t a b o l i t e s . 2 0 2 0 ; 10 : 1 5 8 . h t t p s : / / d o i . o r g / 1 0. 3 3 9 0 / clinical application. Ann Epidemiol. 2009;19:73–8. https://doi. metabo10040158. org/10.1016/j.annepidem.2007.12.001. 134. van der Laan T, Boom I, Maliepaard J, Dubbelman A-C, Harms 150. Murphy BM, He X, Dandy D, Henry CS. Competitive immuno- AC, Hankemeier T. Data-independent acquisition for the quanti- assays for simultaneous detection of metabolites and proteins fication and identification of metabolites in plasma. Metabolites. using micromosaic patterning. Anal Chem. 2008;80:444–50. 2020;10. https://doi.org/10.3390/metabo10120514. https://doi.org/10.1021/ac7019046. 135. Cifuentes Girard MF, Ruskic D, Böhm G, Picenoni R, 151. Ye F, Liu J, Huang Y, Li S, Zhao S. Competitive immunoassay of Hopfgartner G. Automated parallel derivatization of metabolites progesterone by microchip electrophoresis with chemilumines- with SWATH-MS data acquisition for qualitative and quantita- cence detection. J Chromatogr B Analyt Technol Biomed Life tive analysis. Anal Chim Acta. 2020;1127:198–206. https://doi. Sci. 2013;936:74–9. https://doi.org/10.1016/j.jchromb.2013.08. org/10.1016/j.aca.2020.06.030. 002. Castelli F.A. et al. 152. Apilux A, Rengpipat S, Suwanjang W, Chailapakul O. Paper- 165. Watanabe T, Tanaka R, Ono H, Suzuki Y, Tatsuta R, Itoh H. based immunosensor with competitive assay for cortisol detection. Sensitive, wide-range and high-throughput quantification of cy- J Pharm Biomed Anal. 2020;178:112925. https://doi.org/10.1016/ closporine in whole blood using ultra-performance liquid chroma- j.jpba.2019.112925. tography coupled to tandem mass spectrometry and comparison 153. Sanavio B, Krol S. On the slow diffusion of point-of-care systems with an antibody-conjugated magnetic immunoassay. Biomed in therapeutic drug monitoring. Front Bioeng Biotechnol. 2015;3. Chromatogr BMC. 2021;35:e5128. https://doi.org/10.1002/bmc. https://doi.org/10.3389/fbioe.2015.00020. 5128. 154. Reynoso E, Torres E, Bettazzi F, Palchetti I. Trends and perspec- 166. Bahadır EB, Sezgintürk MK (2016) Lateral flow assays: tives in immunosensors for determination of currently-used pesti- Principles, designs and labels. TrAC Trends Anal Chem 82: cides: the case of glyphosate, organophosphates, and 286–306. https://doi.org/10.1016/j.trac.2016.06.006. neonicotinoids. Biosensors. 2019;9:20. https://doi.org/10.3390/ 167. Dincer C, Bruch R, Kling A, Dittrich PS, Urban GA. Multiplexed bios9010020. point-of-care testing – xPOCT. Trends Biotechnol. 2017;35:728– 155. Li Y, Zhang G, Mao X, Yang S, De Ruyck K, Wu Y. High 42. https://doi.org/10.1016/j.tibtech.2017.03.013. sensitivity immunoassays for small molecule compounds detec- 168. Quesada-González D, Merkoçi A. Nanomaterial-based devices for tion – novel noncompetitive immunoassay designs. TrAC Trends point-of-care diagnostic applications. Chem Soc Rev. 2018;47: Anal Chem. 2018;103:198–208. https://doi.org/10.1016/j.trac. 4697–709. https://doi.org/10.1039/C7CS00837F. 2018.04.008. 169. Liang W, Lin H, Chen J, Chen C. Utilization of nanoparticles in 156. Roberts RF, Roberts WL. Performance characteristics of five au- microfluidic systems for optical detection. Microsyst Technol. tomated serum cortisol immunoassays. Clin Biochem. 2004;37: 2016;22:2363–70. https://doi.org/10.1007/s00542-016-2921-4. 489–93. https://doi.org/10.1016/j.clinbiochem.2004.01.009. 170. Fuentes-Chust C, Parolo C, Rosati G, Rivas L, Perez-Toralla K, 157. Wen K, Bai Y, Wei Y, Li C, Shen J, Wang Z. Influence of small Simon S, de Lecuona I, Junot C, Trebicka J, Merkoçi A. The molecular property on antibody response. J Agric Food Chem. microbiome meets nanotechnology: opportunities and challenges 2020;68:10944–50. https://doi.org/10.1021/acs.jafc.0c04333. in developing new diagnostic devices. Adv Mater Deerfield Beach 158. Brozmanova H, Urinovska R, Safarcik K, Vsiansky F, Kacirova I, Fla. 2021;33:e2006104. https://doi.org/10.1002/adma. Grundmann M. Liquid chromatography-tandem mass spectrome- 202006104. try method for quantification of gentamicin and its individual con- 171. Sempionatto JR, Jeerapan I, Krishnan S, Wang J. Wearable chem- geners in serum and comparison results with two ical sensors: emerging systems for on-body analytical chemistry. immunoanalytical methods (fluorescence polarization immunoas- Anal Chem. 2020;92:378–96. https://doi.org/10.1021/acs. say and chemiluminiscent microparticle immunoassay). Clin analchem.9b04668. Chim Acta Int J Clin Chem. 2021;S0009-8981(21):00254. 172. Land KJ, Boeras DI, Chen X-S, Ramsay AR, Peeling RW. https://doi.org/10.1016/j.cca.2021.07.014. Reassured diagnostics to inform disease control strategies, 159. Bowerbank SL, Carlin MG, Dean JR. A direct comparison of strengthen health systems and improve patient outcomes. Nat liquid chromatography-mass spectrometry with clinical routine Microbiol. 2019;4:46–54. https://doi.org/10.1038/s41564-018- testing immunoassay methods for the detection and quantification 0295-3. of thyroid hormones in blood serum. Anal Bioanal Chem. 173. Tuerk C, Gold L. Systematic evolution of ligands by exponential 2019;411:2839–53. https://doi.org/10.1007/s00216-019-01724-2. enrichment: RNA ligands to bacteriophage T4 DNA polymerase. 160. Shin S-Y, Kwon M-J, Song J, Park H, Woo H-Y. Measurement of Science. 1990;249:505–10. https://doi.org/10.1126/science. serum total vitamin D (25-OH) using automated immunoassay in 2200121. comparison [corrected] with liquid chromatography tandem-mass 174. Ellington AD, Szostak JW. In vitro selection of RNA molecules spectrometry. J Clin Lab Anal. 2013;27:284–9. https://doi.org/10. that bind specific ligands. Nature. 1990;346:818–22. https://doi. 1002/jcla.21598. org/10.1038/346818a0. 161. Farrell C-JL, Martin S, McWhinney B, Straub I, Williams P, 175. Stoltenburg R, Nikolaus N, Strehlitz B. Capture-SELEX: selection Herrmann M. State-of-the-art vitamin D assays: a comparison of of DNA aptamers for aminoglycoside antibiotics. J Anal Methods automated immunoassays with liquid chromatography-tandem Chem. 2012;1. https://doi.org/10.1155/2012/415697. mass spectrometry methods. Clin Chem. 2012;58:531–42. 176. Mendonsa SD, Bowser MT. In vitro evolution of functional DNA https://doi.org/10.1373/clinchem.2011.172155. using capillary electrophoresis. J Am Chem Soc. 2004;126:20–1. 162. Moal V, Mathieu E, Reynier P, Malthièry Y, Gallois Y. Low https://doi.org/10.1021/ja037832s. serum testosterone assayed by liquid chromatography-tandem 177. McKeague M, Derosa MC. Challenges and opportunities for small mass spectrometry. Comparison with five immunoassay tech- molecule aptamer development. J. Nucleic Acids. 2012;2012. niques. Clin Chim Acta Int J Clin Chem. 2007;386:12–9. https:// 178. Bauer M, Strom M, Hammond DS, Shigdar S. Anything you can doi.org/10.1016/j.cca.2007.07.013. do, i can do better: can aptamers replace antibodies in clinical 163. Huhtaniemi IT, Tajar A, Lee DM, O’Neill TW, Finn JD, Bartfai G, diagnostic applications? Molecules. 2019;24:1–13. https://doi. Boonen S, Casanueva FF, Giwercman A, Han TS, Kula K, Labrie F, org/10.3390/molecules24234377. Lean MEJ, Pendleton N, Punab M, Silman AJ, Vanderschueren D, 179. Thiviyanathan V, Gorenstein GD. Aptamers and the next genera- Forti G, Wu FCW, EMAS Group. Comparison of serum testosterone tion of diagnostic reagents. Proteomics Clin Appl. 2012;6:563–73. and estradiol measurements in 3174 European men using platform 180. Rampey AM, Umpleby RJ, Rushton GT, Iseman JC, Shah RN, immunoassay and mass spectrometry; relevance for the diagnostics Shimizu KD. Characterization of the imprint effect and the influ- in aging men. Eur J Endocrinol. 2012;166:983–91. https://doi.org/ ence of imprinting conditions on affinity, capacity, and heteroge- 10.1530/EJE-11-1051. neity in molecularly imprinted polymers using the Freundlich 164. Martínez-Escribano A, Maroto-García J, Ruiz-Galdón M, Barrios- isotherm-affinity distribution analysis. Anal Chem. 2004;76: Rodríguez R, Álvarez-Millán JJ, Cabezas-Sánchez P, Plaza- 1123–33. https://doi.org/10.1021/ac0345345. Andrades I, Molina-Vega M, Tinahones FJ, Queipo-Ortuño MI, 181. Spivak D. Optimization, evaluation, and characterization of mo- Fernández-García JC. Measurement of serum testosterone in non- lecularly imprinted polymers. Adv Drug Deliv Rev. 2005;57: diabetic young obese men: comparison of direct immunoassay to 1779–94. https://doi.org/10.1016/j.addr.2005.07.012. liquid chromatography-tandem mass spectrometry. Biomolecules. 182. Naseri M, Mohammadniaei M, Sun Y, Ashley J. The use of 2020;10:E1697. https://doi.org/10.3390/biom10121697. aptamers and molecularly imprinted polymers in biosensors for Metabolomics for personalized medicine: the input of analytical chemistry from biomarker discovery to... environmental monitoring: a tale of two receptors. Chemosensors. 198. Luo Y, Yu H, Alkhamis O, Liu Y, Lou X, Yu B, Xiao Y. Label- 2020;8:32. https://doi.org/10.3390/chemosensors8020032. free, visual detection of small molecules using highly target- 183. Yarman A, Scheller F. The first electrochemical MIP sensor for responsive multimodule split aptamer constructs. Anal Chem. tamoxifen. Sensors. 2014;14:7647–54. https://doi.org/10.3390/ 2019;91:7199–207. https://doi.org/10.1021/acs.analchem. s140507647. 9b00507. 184. Menon S, Rajeshkumar S, Venkat Kumar S. A review on biogenic 199. Swensen JS, Xiao Y, Ferguson BS, Lubin AA, Lai RY, Heeger synthesis of gold nanoparticles, characterization, and its applica- AJ, Plaxco KW, Soh HT. Continuous, real-time monitoring of tions. Resour-Effic Technol. 2017;3:516–27. https://doi.org/10. cocaine in undiluted blood serum via a microfluidic, electrochem- 1016/j.reffit.2017.08.002. ical aptamer-based sensor. J Am Chem Soc. 2009;131:4262–6. 185. Lee S, Lin OH, Doong R. Design of size-tunable molecularly https://doi.org/10.1021/ja806531z. imprinted polymer for selective adsorption of acetaminophen. 200. Chen J, Fang Z, Lie P, Zeng L. Computational lateral flow bio- Clean Technol Environ Policy. 2017;19:243–50. https://doi.org/ sensor for proteins and small molecules: a new class of strip logic 10.1007/s10098-016-1201-4. gates. Anal Chem. 2012;84:6321–5. https://doi.org/10.1021/ 186. Diliën H, Peeters M, Royakkers J, Harings J, Cornelis P, Wagner ac301508b. P, Steen Redeker E, Banks CE, Eersels K, van Grinsven B, Cleij 201. Fu H, Yang J, Guo L, Nie J, Yin Q, Zhang L, Zhang Y. Using the TJ. Label-free detection of small organic molecules by molecular- Rubik’s cube to directly produce paper analytical devices for ly imprinted polymer functionalized thermocouples: toward quantitative point-of-care aptamer-based assays. Biosens in vivo applications. ACS Sens. 2017;2:583–9. https://doi.org/ Bioelectron. 2017;96:194–200. https://doi.org/10.1016/j.bios. 10.1021/acssensors.7b00104. 2017.05.012. 187. Moret J, Moreira FTC, Almeida SAA, Sales MGF. New molecularly- 202. Canoura J, Wang Z, Yu H, Alkhamis O, Fu F, Xiao Y. No imprinted polymer for carnitine and its application as ionophore in structure-switching required: a generalizable exonuclease- potentiometric selective membranes. Mater Sci Eng C. 2014;43:481– mediated aptamer-based assay for small-molecule detection. J 7. https://doi.org/10.1016/j.msec.2014.07.056. Am Chem Soc. 2018;140:9961–71. https://doi.org/10.1021/jacs. 188. Sehit E, Drzazgowska J, Buchenau D, Yesildag C, Lensen M, 8b04975. Altintas Z. Ultrasensitive nonenzymatic electrochemical glucose 203. Lee J-R, Choi J, Shultz TO, Wang SX. Small molecule detection sensor based on gold nanoparticles and molecularly imprinted in saliva facilitates portable tests of marijuana abuse. Anal Chem. polymers. Biosens Bioelectron. 2020;165:112432. https://doi. 2016;88:7457–61. https://doi.org/10.1021/acs.analchem. org/10.1016/j.bios.2020.112432. 6b01688. 189. Zhou W, Jimmy Huang PJ, Ding J, Liu J. Aptamer-based biosen- 204. Park J-H, Byun J-Y, Jang H, Hong D, Kim M-G. A highly sensi- sors for biomedical diagnostics. Analyst. 2014;139:2627–40. tive and widely adaptable plasmonic aptasensor using berberine https://doi.org/10.1039/c4an00132j. for small-molecule detection. Biosens Bioelectron. 2017;97:292– 190. Seok Kim Y, Ahmad Raston NH, Bock Gu M. Aptamer-based 8. https://doi.org/10.1016/j.bios.2017.06.019. nanobiosensors. Biosens Bioelectron. 2016;76:2–19. https://doi. 205. Ruta J, Perrier S, Ravelet C, Fize J, Peyrin E. Noncompetitive org/10.1016/j.bios.2015.06.040. fluorescence polarization aptamer-based assay for small molecule 191. Prante M, Segal E, Scheper T, Bahnemann J, Walter J. detection. Anal Chem. 2009;81:7468–73. https://doi.org/10.1021/ Aptasensors for point-of-care detection of small molecules. ac9014512. Biosensors. 2020;10. https://doi.org/10.3390/bios10090108. 206. Yen Y-K, Chiu C-Y. A CMOS MEMS-based membrane-bridge 192. Zuo X, Xiao Y, Plaxco KW. High specificity, electrochemical nanomechanical sensor for small molecule detection. Sci Rep. sandwich assays based on single aptamer sequences and suitable 2020;10:2931. https://doi.org/10.1038/s41598-020-60057-8. for the direct detection of small-molecule targets in □ blood and 207. Yang J, Hu Y, Li Y. Molecularly imprinted polymer-decorated other complex matrices. J Am Chem Soc. 2009;131:6944–5. signal on-off ratiometric electrochemical sensor for selective and https://doi.org/10.1021/ja901315w. robust dopamine detection. Biosens Bioelectron. 2019;135:224– 193. Alyamani BJ, Alsager OA, Zourob M. Label-free fluorescent 30. https://doi.org/10.1016/j.bios.2019.03.054. aptasensor for small targets via displacement of groove bound 208. Tsugawa H, Cajka T, Kind T, Ma Y, Higgins B, Ikeda K, curcumin molecules. Sens Switz. 2019;19. https://doi.org/10. Kanazawa M, VanderGheynst J, Fiehn O, Arita M. MS-DIAL: 3390/s19194181. data-independent MS/MS deconvolution for comprehensive me- 194. D’Alessandro A. Benford’s law and metabolomics: a tale of num- tabolome analysis. Nat Methods. 2015;12:523–6. https://doi.org/ bers and blood. Transfus Apher Sci Off J World Apher Assoc Off 10.1038/nmeth.3393. J Eur Soc Haemapheresis. 2020;59:103019. https://doi.org/10. 1016/j.transci.2020.103019. 209. Kim K, Jo E-J, Lee KJ, Park J, Jung GY, Shin Y-B, Lee LP, Kim 195. Hsu Y-HH, Astley CM, Cole JB, Vedantam S, Mercader JM, M-G. Gold nanocap-supported upconversion nanoparticles for Metspalu A, Fischer K, Fortney K, Morgen EK, Gonzalez C, fabrication of a solid-phase aptasensor to detect ochratoxin A. Gonzalez ME, Esko T. Hirschhorn JN (2020) Integrating untargeted Biosens Bioelectron. 2020;150:111885. https://doi.org/10.1016/j. metabolomics, genetically informed causal inference, and pathway bios.2019.111885. enrichment to define the obesity metabolome. Int J Obes. 2005;44: 210. Idili A, Parolo C, Ortega G, Plaxco KW. Calibration-free measure- 1596–606. https://doi.org/10.1038/s41366-020-0603-x. ment of phenylalanine levels in the blood using an electrochemical 196. Wang Y, Hussein AM, Somasundaram L, Sankar R, Detraux D, aptamer-based sensor suitable for point-of-care applications. ACS Mathieu J, Ruohola-Baker H. microRNAs regulating human and Sens. 2019;4:3227–33. https://doi.org/10.1021/acssensors. mouse naïve pluripotency. Int J Mol Sci. 2019;20:5864. https:// 9b01703. doi.org/10.3390/ijms20235864. 211. Wu S, Liu L, Duan N, Li Q, Zhou Y, Wang Z. Aptamer-based 197. Goodacre R, Broadhurst D, Smilde AK, Kristal BS, Baker JD, lateral flow test strip for rapid detection of zearalenone in corn Beger R, Bessant C, Connor S, Capuani G, Craig A, Ebbels T, samples. J Agric Food Chem. 2018;66:1949–54. https://doi.org/ Kell DB, Manetti C, Newton J, Paternostro G, Somorjai R, 10.1021/acs.jafc.7b05326. Sjöström M, Trygg J, Wulfert F. Proposed minimum reporting standards for data analysis in metabolomics. Metabolomics. Publisher’s note Springer Nature remains neutral with regard to jurisdic- 2007;3:231–41. https://doi.org/10.1007/s11306-007-0081-3. tional claims in published maps and institutional affiliations.