Validation of Pooled Whole-Genome Re-Sequencing in Arabidopsis lyrata

• Published in: PloS one Vol. 10; no. 10; p. e0140462
• Main Authors: Fracassetti, Marco, Griffin, Philippa C, Willi, Yvonne
• Format: Journal Article
• Language: English
• Published: United States Public Library of Science 13-10-2015; Public Library of Science (PLoS)
• Subjects: Accuracy; Analysis; Arabidopsis; Arabidopsis - genetics; Base sequence; Bayesian analysis; Bioinformatics; Biology; Correlation; Correlation coefficient; Correlation coefficients; Data acquisition; Deoxyribonucleic acid; DNA; DNA sequencing; Estimates; Frequency estimation; Gene frequency; Gene Frequency - genetics; Gene sequencing; Genetic aspects; Genetic testing; Genome, Plant; Genomes; Genomics; Genotyping; Internet; Parks & recreation areas; Plant sciences; Polymorphism; Polymorphism, Single Nucleotide - genetics; Population genetics; Reproducibility of Results; Sequence Analysis, DNA - methods; Single nucleotide polymorphisms; Single-nucleotide polymorphism; Statistics as Topic; United States > US; Switzerland
• ISSN: 1932-6203; 1932-6203
• DOI: 10.1371/journal.pone.0140462

Sequencing pooled DNA of multiple individuals from a population instead of sequencing individuals separately has become popular due to its cost-effectiveness and simple wet-lab protocol, although some criticism of this approach remains. Here we validated a protocol for pooled whole-genome re-sequencing (Pool-seq) of Arabidopsis lyrata libraries prepared with low amounts of DNA (1.6 ng per individual). The validation was based on comparing single nucleotide polymorphism (SNP) frequencies obtained by pooling with those obtained by individual-based Genotyping By Sequencing (GBS). Furthermore, we investigated the effect of sample number, sequencing depth per individual and variant caller on population SNP frequency estimates. For Pool-seq data, we compared frequency estimates from two SNP callers, VarScan and Snape; the former employs a frequentist SNP calling approach while the latter uses a Bayesian approach. Results revealed concordance correlation coefficients well above 0.8, confirming that Pool-seq is a valid method for acquiring population-level SNP frequency data. Higher accuracy was achieved by pooling more samples (25 compared to 14) and working with higher sequencing depth (4.1× per individual compared to 1.4× per individual), which increased the concordance correlation coefficient to 0.955. The Bayesian-based SNP caller produced somewhat higher concordance correlation coefficients, particularly at low sequencing depth. We recommend pooling at least 25 individuals combined with sequencing at a depth of 100× to produce satisfactory frequency estimates for common SNPs (minor allele frequency above 0.05).