Bioinformatics, 36(7), 2020, 2298–2299 doi: 10.1093/bioinformatics/btz862 Advance Access Publication Date: 18 November 2019 Applications Note Systems biology Link-HD: a versatile framework to explore and integrate heterogeneous microbial communities Laura M. Zingaretti 1,2,*, Gilles Renand3, Diego P. Morgavi4 and Yuliaxis Ramayo-Caldas3,5 1 Plant and Animal Genomics, Statistical and Population Genomics Group, CSIC-IRTA-UAB-UB Consortium, Centre for Research in Agricultural Genomics (CRAG), 08193 Bellaterra, Spain, 2IAPCBA and IAPCH, UNVM, Villa Marı́a, Córdoba 5900, Argentina, 3URM Animal Genetics and Integrative Biology, GABI, INRA, AgroParisTech, Université Paris-Saclay, 78352 Jouy-en-Josas, France, 4Animal Physiology and Livestock Systems Divisions, INRA, Herbivore Research Unit, Clermont Auvergne University, Saint Genès-Champanelle 63122, France and 5Animal Breeding and Genetics Program, IRTA, 08140 Caldes de Montbui, Spain *To whom correspondence should be addressed. Associate Editor: Alfonso Valencia Received on June 19, 2019; revised on October 4, 2019; editorial decision on November 14, 2019; accepted on November 15, 2019 Abstract Motivation: We present Link-HD, an approach to integrate multiple datasets. Link-HD is a generalization of ‘Structuration des Tableaux A Trois Indices de la Statistique–Analyse Conjointe de Tableaux’, a family of methods designed to integrate information from heterogeneous data. Here, we extend the classical approach to deal with broader datasets (e.g. compositional data), methods for variable selection and taxon-set enrichment analysis. Results: The methodology is demonstrated by integrating rumen microbial communities from cows for which me- thane yield (CH4 y ) was individually measured. Our approach reproduces the significant link between rumen micro- biota structure and CH4 emission. When analyzing the TARA’s ocean data, Link-HD replicates published results, highlighting the relevance of temperature with members of phyla Proteobacteria on the structure and functionality of this ecosystem. Contact: m.lau.zingaretti@gmail.com Availability and implementation: The source code, examples and a complete manual are freely available in GitHub https://github.com/lauzingaretti/LinkHD and in Bioconductor https://bioconductor.org/packages/release/bioc/html/ LinkHD.html. 1 Introduction 1976), a family of multivariate methods to integrate multiple datasets. Link-HD generalizes STATIS with Regression Biplot The reduction of ‘omics’ technology costs now enables collection (Ter Braak, 1997), clustering, differential abundance, enrichment of data from multiple sources. This allows researchers to simultan- taxonomic analysis and visualization tools. Link-HD analyzes eously study several datasets and investigate their relationship distance tables computed from numerical, categorical, or com- with complex traits. The integration of these heterogeneous data- positional data as a generalization of multidimensional scaling sets is not trivial and several statistical methods have been devel- (Abdi et al., 2007). Furthermore, Link-HD performs variable se- oped to address this challenge (Argelaguet et al., 2018; Mariette lection and can link the obtained common sub-space with pheno- and Villa-Vialaneix, 2018; Meng et al., 2014). In particular, the type information. amalgamation of multiple microbial ecosystems poses unique chal- lenges as these are compositional and sparse data. MixKernel (Mariette and Villa-Vialaneix, 2018) is a well-known tool designed to integrate heterogeneous datasets including microbial 2 Materials and methods communities, but no method to perform a taxonomic enrichment Like STATIS, Link-HD aims to compare and analyze the relation- analysis is available. Another popular integrative approach is ships between datasets with a shared set of observations or varia- MOFA (Argelaguet et al., 2018), however, it is unable to deal with bles. However, our package was specifically designed to integrate compositional data. microbial communities and incorporate distances and transforma- Here, we present Link-HD, a tool to integrate and explore tions to deal with compositional data (Aitchison, 1982). The multiple microbial communities based on STATIS (Des Plantes, method is implemented in three main phases (Fig. 1). C The Author(s) 2019. Published by Oxford University Press. V 2298 This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact journals.permissions@oup.com Link-HD 2299 replicates the relevant role of temperature and Proteobacteria phyla on the structure of this ecosystem, as described in Mariette and Villa-Vialaneix (2018). Finally, we show the potential of Link-HD to integrate other omics layers by using transcriptome NCI-60 cell lines. Link-HD recapitulates the reported data structure (Meng et al., 2014) and ontology analysis reveals several cancer-related pathways. In all, our results demonstrate that Link-HD is robust in combin- ing several heterogeneous data types. A detailed description of these case studies and the theory behind Link-HD is available at https:// lauzingaretti.github.io/LinkHD/ and in Bioconductor (https://biocon ductor.org/packages/release/bioc/html/LinkHD.html). 4 Conclusions We have developed an R package to integrate multiple microbial communities and other ‘omics’ layers combining a plethora of statis- tical methods in a fast, simple and flexible way. Acknowledgements Fig. 1. Link-HD Workflow. In the Inter-structure step, raw data are transformed The authors would like to thank Miguel Pérez-Enciso, Rayner González- using cumulative sum scaling or centered log ratio, and the correlation coefficient Prendes and Jordi Estelle for valuable discussions and comments on the manu- (Rv) is computed. The second step is the compromise (W) and, finally, the intra- script and also to APIS-GENE for funding the MicroFicient project. structure step involves the Eigen-decomposition of W. Observations can be clustered and methods for selecting variables and association with phenotypes are available Funding i. Inter-structure step: The algorithm first assesses the similarity L.M.Z. is recipient of a Ph.D. grant from Ministry of Economy and Science, between transformed distance tables using the vector correl- Spain associated with ‘Centro de Excelencia Severo Ochoa 2016–2019’ award ation coefficient (Rv) (Escoufier, 1973), which can be inter- SEV-2015-0533 to CRAG. Y.R.C. was funded by Marie Skłodowska-Curie preted as a general ‘vector covariance’ between matrices, i.e. grant (P-Sphere) agreement No 6655919 (EU). this step evaluates similarity between the disparate datasets. Conflict of Interest: none declared. ii. Compromise step: Next, the ‘compromise’ matrix is calculated, which is a weighted sum of each cross-product matrix. This step involves an optimization problem since the weights are chosen References to maximize the correlation between the compromise matrix Abdi,H. et al. (2007) Analyzing assessors and products in sorting tasks: and each individual component. DISTATIS, theory and applications. Food Qual. Prefer., 18, 627–640. Aitchison,J. (1982) The statistical analysis of compositional data. J. R. Stat. iii. Intra-structure step: Finally, the compromise matrix is eval- Soc. Ser. B, 44, 139–160. uated through a Principal Component Analysis. The coordi- Argelaguet,R. et al. (2018) Multi-Omics Factor Analysis—a framework for nates of the common elements are projected into a low rank unsupervised integration of multi-omics data sets. Mol. Syst. Biol., 14, space, where the relationships between them can be easily e8124. interpreted. Danielsson,R. et al. (2017) Methane production in dairy cows correlates with rumen methanogenic and bacterial community structure. Front. Microbiol., Variable selection is tackled by two alternative approaches: (i) 8, 226. by projecting all the input variables into the compromise through a Des Plantes,H.L. (1976) Structuration des tableaux à trois indices de la statis- general Biplot formulation (Ter Braak, 1997); and (ii) by computing tique: théorie et application d’une méthode d’analyse conjointe. Doctoral dissertation, Université des sciences et techniques du Languedoc. the differential abundance of features between clusters of samples. A Escoufier,Y. (1973) Le traitement des variables vectorielles. Biometrics, 29, novelty of Link-HD is its ability to aggregate the selected variables 751–760. at several taxonomic levels and to establish whether that level is Kittelmann,S. et al. (2014) Two different bacterial community types are linked enriched using a cumulative hypergeometric distribution. This func- with the low-methane emission trait in sheep. PLoS One, 9, e103171. tion also allows users to add a custom OTUs list. Finally, the SPIEC- Kurtz,Z.D. et al. (2015) Sparse and compositionally robust inference of micro- EASI (Kurtz et al., 2015) tool can be used to visualize variable bial ecological networks. PLoS Comput. Biol., 11, e1004226. interactions. Mariette,J. and Villa-Vialaneix,N. (2018) Unsupervised multiple kernel learn- ing for heterogeneous data integration. Bioinformatics, 34, 1009–1015. Meng,C. et al. (2014) A multivariate approach to the integration of 3 Case studies multi-omics datasets. BMC Bioinformatics, 15, 162. Ramayo-Caldas,Y. et al. (2019) Identification of rumen microbial biomarkers We illustrate our approach with rumen microbial (Ramayo-Caldas linked to methane emission in Holstein dairy cows. J. Anim. Breed. Genet., et al., 2019), TARA’s Ocean expedition (Sunagawa et al., 2015) and doi: 10.1111/jbg.12427. transcriptome NCI-60 cell line datasets (Reinhold et al., 2012). Reinhold,W.C. et al. (2012) CellMiner: a web-based suite of genomic and In the rumen study, we integrated Bacteria, Archaea and pharmacologic tools to explore transcript and drug patterns in the NCI-60 Protozoa from 65 Holstein cows. Link-HD was able to reproduce cell line set. Cancer Res., 72, 3499–3511. previous results (Danielsson et al., 2017; Kittelmann et al., 2014; Sunagawa,S. et al. (2015) Ocean plankton. Structure and function of the glo- Ramayo-Caldas et al., 2019), showing a link between the structure bal ocean microbiome. Science, 348, 1261359. of the rumen microbiota and CH4 emission. We also identify micro- Ter Braak,C.J.F. (1997) JC Gower and DJ Hand, biplots. Monographs on sta- bial markers associated to CH4. In the TARA’s example, Link-HD tistics and applied probability. Psychometrika, 62, 457–460.