HARSH

HARSH infers haplotypes by integrating multi-SNP short read sequencing data with reference haplotype panels to improve phasing accuracy for genetic linkage and variation analyses.

Key Features:

• Integration of multi-SNP reads: Incorporates information from short reads that span multiple SNPs to capture read-level haplotypic origin data.
• Reference-panel conditioning: Conditions haplotype inference on a reference haplotype panel to identify explanatory haplotype segments.
• Probabilistic model: Uses a novel probabilistic model to identify the most likely haplotype segments from the reference panel that explain an individual's sequencing data.
• Gibbs sampler: Employs an efficient Gibbs sampler-based sampling method to perform inference.
• Relation to HMM/copying models: Builds on the haplotype copying/HMM framework for representing individual haplotypes as mosaics of reference haplotypes.
• Empirical performance: Demonstrates approximately 20% higher accuracy versus basic haplotype copying models and about 10% higher accuracy versus Hap-SeqX, with reduced computational time and memory.
• Benchmark datasets: Validated using simulated sequencing reads derived from real genotypes in the HapMap project and the 1000 Genomes Project.

Scientific Applications:

• Haplotype phasing from sequencing: Phasing of individual genomes using short read sequencing data combined with reference panels.
• Improved genotype calling and phasing accuracy: Enhancing accuracy of genotype calls and phased haplotypes by leveraging multi-SNP read information.
• Population genetics analyses: Facilitating analyses of genetic linkage and variation using HapMap and 1000 Genomes reference panels.

Methodology:

Implements a probabilistic model that integrates multi-SNP short read data and conditions on a reference haplotype panel to identify likely reference haplotype segments using an efficient Gibbs sampler.