QCALL

QCALL detects and genotypes single-nucleotide polymorphisms (SNPs) from low-coverage sequencing data of multiple diploid samples by leveraging shared haplotypes, ancestral recombination graphs (ARGs), and population-genetic models to improve SNP-calling accuracy and phased genotype inference.

Key Features:

• Low-Coverage Sequencing Efficiency: Enables SNP detection and genotyping from low-coverage whole-genome sequencing across large sample sets.
• Shared Haplotype Utilization: Uses linkage disequilibrium and shared haplotype information across samples to enhance SNP candidate detection and genotyping accuracy.
• SNP Candidate Collection: Identifies SNP candidates based on independent sequence calls at each site across the population samples.
• Ancestral Recombination Graphs (ARGs) via MARGARITA: Employs MARGARITA to generate 20 ARGs from genotype or phased haplotype data and refines SNP probabilities by considering mutations on internal branches of inferred marginal ancestral trees.
• Bayesian Inference and Population-Genetic Priors: Integrates a population-genetic prior on tree-branch lengths with Bayesian inference to compute posterior probabilities that SNPs are real and to infer the most probable phased genotypes per individual.
• Tradeoff Analysis: Analyzes the balance between sequencing depth and sample size to inform resource allocation for SNP detection.
• Experimental Validation: Methods have been validated using simulated data and real datasets from the 1000 Genomes Project.

Scientific Applications:

• Population Genetics Studies: Identification and genotyping of genetic variants within populations to study evolutionary dynamics and population structure.
• Large-Scale Genomic Projects: Enabling variant discovery and phasing in studies where extensive sample sizes are prioritized over high per-sample sequencing depth.

Methodology:

SNP candidates are collected from independent sequence calls at each site across samples. MARGARITA is used to construct 20 ancestral recombination graphs from genotype or phased haplotype data, and mutations along internal branches of inferred marginal trees are analyzed to refine SNP candidate probabilities. A population-genetic prior on tree-branch lengths combined with Bayesian inference computes posterior probabilities for SNPs and determines the most probable phased genotype per individual. The approach includes analysis of the tradeoff between sequencing depth and sample size.