Benchmarking the performance of Pool-seq SNP callers using simulated and real sequencing data

Area of Science:

• Population genomics
• Bioinformatics
• Genetics

Background:

• Population genomics leverages genome-wide data to identify molecular variants and evolutionary forces.
• High costs of individual genome sequencing limit large-scale population studies.
• Pooled sequencing (Pool-seq) offers a cost-effective alternative for detecting single nucleotide polymorphisms (SNPs) and estimating allele frequencies.

Purpose of the Study:

• To benchmark the performance of different SNP callers for Pool-seq data.
• To evaluate SNP caller performance under various conditions using simulated and real data.
• To identify optimal SNP calling strategies for population genomics studies.

Main Methods:

• Benchmarking of SNP callers (SNAPE-pooled, MAPGD, VarScan, PoolSNP) for Pool-seq data.
• Utilized computer simulations and real sequencing data for performance evaluation.
• Assessed performance based on sensitivity and false discovery rate (FDR) across different allele frequencies.

Main Results:

• SNP caller performance varied significantly, particularly for allele frequencies up to 0.35.
• Bayesian (SNAPE-pooled) and maximum likelihood (MAPGD) callers outperformed heuristic callers (VarScan, PoolSNP).
• Bayesian and maximum likelihood approaches demonstrated a better balance between sensitivity and FDR in both simulated and real data.

Conclusions:

• The choice of SNP caller is critical for successful population genomics using Pool-seq.
• Bayesian and maximum likelihood methods are recommended for accurate SNP calling in Pool-seq data.
• These findings are valuable for large-scale population studies, metagenomics, and polyploid research.