Conservation Genetics Resources (2023) 15:113–116 https://doi.org/10.1007/s12686-023-01309-3 METHODS AND RESOURCES SNP+ to predict dropout rates in SNP arrays N. Sastre1 · A. Mercadé1 · J. Casellas2 Received: 14 November 2022 / Accepted: 12 June 2023 / Published online: 8 July 2023 © The Author(s) 2023 Abstract Genotyping individuals using forensic or non-invasive samples such as hair or fecal samples increases the risk of allelic amplification failure (dropout) due to the low quality and quantity of DNA. One way to decrease genotyping errors is to increase the number of replicates per sample. Here, we have developed the software SNP+ to estimate the dropout prob- ability and the subsequent required number of replicates to obtain the reliable genotype with probability 95%. Moreover, the software predicts the minor allele frequency and compares two competing models assuming equal or allele-specific dropout probabilities by Bayes factor. The software handles data from one SNP to high density arrays (e.g., 100,000 SNPs). Keywords SNP · Software · Dropout · Forensic samples Introduction reduce genotyping errors using non-invasive samples without cost, we decided to develop a software (SNP+) to predict the Single nucleotide polymorphisms (SNPs) are biallelic mark- dropout probability of each SNP from a sample of replicated ers largely abundant in most genomes and with low mutation genotypes. Moreover, two alternative parametrizations were rate (~ ­10−9 per generation) (Brumfield et al. 2003; Morin compared by a Bayes factor to check for within-SNP homo- et al. 2004). SNPs can be associated to diseases, susceptibility geneous dropout probability against different dropout prob- to environmental factors or quantitative trait locus (Erichsen abilities for each allele. and Chanock 2004; Amos et al. 2008; Casellas et al. 2008; Nickels et al. 2013). In forensic medicine, SNPs can be useful to identify individuals from non-invasive samples by using Material and methods short amplicons (Sobrino et al. 2005). However, individual identification using degraded samples with less than 100 cop- The SNP+ software analyzes each SNP independently, tak- ies of gDNA may cause genotyping errors (Giardina et al. ing as a starting point a vector y of n genotypes ordered by 2009; von Thaden et al. 2020). Allelic amplification failure, individual (m) and replicates within individual (y’ = [y’1 y’2 or dropout is the most common error caused by stochastic … y’m]), where n­ 1 is the number of replicates for the first effects of the PCR reaction (Taberlet and Luikart 1999). To individual, and n = n1 + n2 + … + nm. Assuming two alleles, reduce dropout ratio, multiplex pre-amplification or increased A and B, the Bayesian joint posterior distribution general- replicates per sample could be performed (Bellemain and izes to Taberlet 2004; Sastre et al. 2009). However, both solutions ( ) increase time and cost for genotyping individuals. In order to p(fA , 𝜀A , 𝜀B |𝐲) ∼ p(𝐲|fA , 𝜀A , 𝜀B )p fA p(𝜀A )p(𝜀B ), and focuses on estimating the allele frequency (fA), as well * J. Casellas as the dropout probability for allele A (εA) or B (εB). Taking joaquim.casellas@uab.cat a particular genotype ­yi with possible outcomes AA, AB, N. Sastre BB and missing genotype (miss.), its Bayesian likelihood natalia.sastre@uab.cat is computed as 1 Servei Veterinari de Genètica Molecular, Universitat Autònoma de Barcelona, 08193 Bellaterra, Spain p(yi = AA|fA , 𝜀A , 𝜀B ) = p(AA|AA)p(AA) + p(AA|AB)p(AB) 2 Departament de Ciència Animal i dels Aliments, Universitat Autònoma de Barcelona, 08193 Bellaterra, Spain 13 Vol.:(0123456789) 114 Conservation Genetics Resources (2023) 15:113–116 p(yi = AB|fA , 𝜀A , 𝜀B ) = p(AB|AB)p(AB) For each SNP, the model was solved by a Metropo- lis–Hastings sampling process (Metropolis et al. 1953) with 500,000 iterations after a burn-in period of 10,000 iterations. Two alternative parameterizations (𝜀A = 𝜀B vs.𝜀A ≠ 𝜀B) were p(yi = BB|fA , 𝜀A , 𝜀B ) = p(BB|AB)p(AB) + p(BB|BB)p(BB) compared by Bayes factor (Kass and Raftery 1995). The minimum number of within-individual replicates required p(yi = miss.|fA , 𝜀A , 𝜀B ) = p(miss.|AA)p(AA) to predict the reliable genotype with probability 95% was + p(miss.|AB)p(AB) + p(miss.|BB)p(BB), calculated as log(0.05)/log(εA). All these procedures have been implemented in the SNP+ software, available at http://​ where www.​casel​las.​info/​softw​are.​html. p(AA|AA) = (1−𝜀A )2 Results and discussion p(AA|AB) = 𝜀B The program generates the following text delimited output p(AB|AB) = (1−𝜀A )(1−𝜀B ) files: p(BB|AB) = 𝜀A (1) Summary table of the probability of error, confidence interval, replications, and Bayes factor of all the SNPs (output file example in Fig. 1). p(BB|BB) = (1−𝜀B )2 (2) SNP-by-SNP report of dropout probabilities with their confidence intervals, minimum number of replicates, p(miss.|AA) = 𝜀2A Bayer factor comparing a single dropout probability against two independent dropout probabilities. (3) Predicted genotypes for each individual and SNP, and p(miss.|AB) = 𝜀A 𝜀B probability of error, if any. (4) Pairwise comparison between individuals and the prob- p(miss.|BB) = 𝜀2B ability to have an identical genotype. ( ) (5) SNP-by-SNP report of the minor allele frequency And p(AA) = fA2 , p(AB) = 2fA 1 − fA , and p(BB) = fB2 . (MAF) and probability of identity (PI). Note that the model assumes that a BB individual cannot be genotyped as AA (the probability of false alleles is zero). The software has been extensively tested on simulated A priori distributions for fA, εA and εB were assumed flat data with appealing results. Figure 2 illustrated SNP+ ability between 0 and 1. to detect allele-specific departures in dropout probabilities, Fig. 1  Output file generated by the SNP+ software for each analyzed and the Bayes factor comparing the models with allele-specific and SNP, its alleles, allele-specific or joint dropout probability (and mini- joint dropout probability mum number of replicates to guarantee a 95% genotype probability), 13 Conservation Genetics Resources (2023) 15:113–116 115 as well as the increase in statistical relevance (i.e., Bayes fac- tor) as estimated dropout differences increase. The U-shaped scatter plot misidentified a common dropout probability in less than 3% of the simulated data sets, and this percentage reduced below 1% with seven replicates per sample (results not shown). Sample size does not modify the dropout prob- ability but the accuracy of the (dropout probability) estimate we obtain with SNP+. Moreover, we have used SNP+ to evaluate two panels using Open Array® technology (Thermo Fisher Scientific Inc). We analyzed 22 fecal samples and 114 hair samples from Iberian brown bears (Ursus arctos) using first a 120 Fig. 2  Predicted dropout probabilities on simulated data sets with 50 SNP panel (data prepared for publication but not submitted). individuals and 5 replicate genotypes per individual. Genotypes were To decrease the cost of the analysis, we selected 60 SNPs simulated under Hardy–Weinberg equilibrium (0.5 allele frequency) and dropout probabilities for each allele were 0.1 and 0.2, respec- out of 120 SNPs with the lowest dropout probabilities, and tively. The Bayes factor compared the model with allele-specific we repeated our analysis with SNP+ using 164 fecal sam- dropout probabilities against the same dropout probability for both ples and 173 hair samples. All samples were replicated four alleles times, and about 25% and 20% of low-quality DNA fecal Study A: 22 fecal samples; 120 SNPs Study B: 114 hair samples; 120 SNPs 60 60 50 50 40 40 30 30 20 20 10 10 0 X 0 X 0.00 0.04 0.08 0.12 0.16 0.2 0.24 0.28 0.32 0.36 0.4 0.44 0.48 0.52 0.56 0.6 0.64 0.68 0.72 0.76 0.8 0.00 0.04 0.08 0.12 0.16 0.2 0.24 0.28 0.32 0.36 0.4 0.44 0.48 0.52 0.56 0.6 0.64 0.68 0.72 0.76 0.8 Dropout probability Dropout probability Study C: 164 fecal samples; 60 SNPs Study D: 173 hair samples; 60 SNPs 60 60 50 50 40 40 30 30 20 20 10 10 0 X 0 X 0.00 0.04 0.08 0.12 0.16 0.2 0.24 0.28 0.32 0.36 0.4 0.44 0.48 0.52 0.56 0.6 0.64 0.68 0.72 0.76 0.8 0.00 0.04 0.08 0.12 0.16 0.2 0.24 0.28 0.32 0.36 0.4 0.44 0.48 0.52 0.56 0.6 0.64 0.68 0.72 0.76 0.8 Dropout probability Dropout probability Fig. 3  Histograms showing the relative frequencies (%) of dropout probability in bear samples using the SNP+ software in four cases (“X” = average dropout probability) 13 116 Conservation Genetics Resources (2023) 15:113–116 and hair samples, respectively (call rate < 25%) were not References included in both analyses. Figure 3 shows the relative fre- quencies of the dropout probability for the four studies. The Amos CI, Wu X, Broderick P et al (2008) Genome-wide association dropout probability was clearly low after selecting the panel scan of tag SNPs identifies a susceptibility locus for lung cancer at 15q25.1. Nat Genet 40(5):616–622. https://​doi.​org/​10.​1038/​ of 60 SNPs using two types of non-invasive samples, on ng.​109 average 0.05 (studies C and D; 60 SNPs) versus 0.2 (studies Bellemain E, Taberlet P (2004) Improved noninvasive genotyping A and B; 120 SNPs). In terms of variability and distribution method: application to brown bear (Ursus arctos) faeces. Mol mode, the study that obtains lower dropout probabilities is Ecol Notes 4(3):519–522. https://​doi.​org/​10.​1111/j.​1471-​8286.​ 2004.​00711.x the study C, after SNP selection. The study with the highest Brumfield R, Beerli PA, Nickerson D et al (2003) The utility of single probability of dropout is the study B probably because hair nucleotide polymorphisms in inferences of population history. samples were hair-trapping collected and therefore, not all Trends Ecol Evol 18:249–256. https://​doi.​org/​10.​1016/​S0169-​ samples contained roots or enough hair quantity to obtain 5347(03)​00018-1 Casellas J, Varona L, Muñoz G et al (2008) Empirical Bayes factor high DNA quality. To summarize, SNP+ calculates the drop- analyses of quantitative trait loci for gestation length in Iberian × out likelihood, the Bayes factor, PI and MAF, and can be Meishan F2 sows. Animal 2(2):177–183. https://​doi.​org/​10.​1017/​ used to select the best arrays from low density arrays up to S1751​73110​70010​85 high density arrays, avoiding those SNPs that require many Erichsen HC, Chanock SJ (2004) SNPs in cancer research and treat- ment. Br J Cancer 90(4):747–751. https://​doi.​org/​10.​1038/​sj.​bjc.​ replicates because they lead to error. Moreover, SNP+ shows 66015​74 the number of replicates needed per sample to reach a 95% Giardina E, Pietrangeli I, Martone C et al (2009) Whole genome ampli- of genotyping reliability per SNP. fication and real-time PCR in forensic casework. BMC Genomics 10:159. https://​doi.​org/​10.​1186/​1471-​2164-​10-​159 Acknowledgements This work was funded by the LoupO project Kass RE, Raftery AE (1995) Bayes factors. J Am Stat Assoc (EFA354/19) of the European Interreg Program V-A Spain-France- 90(430):773–795. https://​doi.​org/​10.​1080/​01621​459.​1995.​10476​ Andorra (POCTEFA 2014-2020). We are grateful to two anonymous 572 reviewers for helpful comments on the manuscript. Metropolis N, Rosenbluth AW, Rosenbluth MN et al (1953) Equation of state calculations by fast computing machines. J Chem Phys Author contributions Conceptualisation: NS, JC, AM. Formal analy- 21(6):1087–1092. https://​doi.​org/​10.​1063/1.​16991​14 sis: JC. Methodology: NS, JC. Resources: NS. Validation: NS, JC. Morin P, Luikart G, Wayne RK (2004) SNPs in ecology, evolution and Writing—original draft: NS Writing—review and editing: NS, JC, AM. conservation. Trends Ecol Evol 19:208–216. https://​doi.​org/​10.​ 1016/j.​tree.​2004.​01.​009 Funding Open Access Funding provided by Universitat Autonoma de Nickels S, Truong T, Hein R et al (2013) Evidence of gene-environ- Barcelona. Funding was provided by Interreg (EFA354/19). ment interactions between common breast cancer susceptibil- ity loci and established environmental risk factors. PLoS Genet Declarations 9(3):e1003284. https://​doi.​org/​10.​1371/​journ​al.​pgen.​10032​84 Sastre N, Francino O, Lampreave G et al (2009) Sex identification of Conflict of interest The authors declare that they have no conflict of wolf (Canis lupus) using non-invasive samples. Conserv Genet interest. 10(3):555–558. https://​doi.​org/​10.​1007/​s10592-​008-​9565-6 Sobrino B, Brión M, Carracedo A (2005) SNPs in forensic genetics: Ethical approval For the present study no animal was captured or a review on SNP typing methodologies. Forensic Sci Int 154(2– euthanized. 3):181–194. https://​doi.​org/​10.​1016/j.​forsc​iint.​2004.​10.​020 Taberlet P, Luikart G (1999) Non-invasive genetic sampling and indi- vidual identification. Biol J Lin Soc 68(1–2):41–55. https://​doi.​ Open Access This article is licensed under a Creative Commons Attri- org/​10.​1111/j.​1095-​8312.​1999.​tb011​57.x bution 4.0 International License, which permits use, sharing, adapta- von Thaden A, Nowak C, Tiesmeyer A et al (2020) Applying genomic tion, distribution and reproduction in any medium or format, as long data in wildlife monitoring: development guidelines for genotyp- as you give appropriate credit to the original author(s) and the source, ing degraded samples with reduced single nucleotide polymor- provide a link to the Creative Commons licence, and indicate if changes phism (SNP) panels. Mol Ecol Resour 20(3):662. https://​doi.​org/​ were made. The images or other third party material in this article are 10.​1111/​1755-​0998.​13136 included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in Publisher's Note Springer Nature remains neutral with regard to the article's Creative Commons licence and your intended use is not jurisdictional claims in published maps and institutional affiliations. permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://​creat​iveco​mmons.​org/​licen​ses/​by/4.​0/. 13