ShoRAH

ShoRAH reconstructs haplotypes and detects low-frequency variants from next-generation sequencing data to characterize genetic diversity in mixed samples while accounting for sequencing errors.

Key Features:

• Error Correction: Clusters reads in small windows to perform error correction and reduce base substitution rates in pyrosequencing HIV-1 data (e.g., from 0.25% to 0.05% and from 0.05% to 0.03%).
• Probabilistic Bayesian Approach: Applies a probabilistic Bayesian framework to distinguish true genetic variants from sequencing errors, improving precision and recall relative to counting-based methods.
• Haplotype Inference: Reconstructs global haplotypes by assigning observed reads to unobserved haplotypes using a generative probabilistic model with a Dirichlet process mixture prior and a Gibbs sampler for local haplotype estimation.
• Strand Bias Testing: Integrates a beta-binomial strand bias test on forward read distributions to reduce false-positive single nucleotide variants (SNVs).
• Application to Viral Populations: Targets genetically diverse viral populations such as HIV and hepatitis C virus to identify low-frequency variants and investigate phenotypic drug resistance and epidemiological links.

Scientific Applications:

• Viral population analysis: Inference of haplotypes and variant spectra from deep sequencing of HIV and hepatitis C virus samples.
• Drug resistance characterization: Detection of minority variants relevant for phenotypic drug resistance analysis.
• Epidemiological linkage: Reconstruction of haplotypes to support analysis of epidemiological links between samples.
• Population genetics and evolution: Study of viral evolution and population genetic structure in heterogeneous sequencing data.

Methodology:

Methods explicitly include read clustering in small windows for error correction, statistical modeling and Bayesian inference, a generative probabilistic model with a Dirichlet process mixture prior, Gibbs sampling for local haplotype estimation, and a beta-binomial strand bias test for SNV filtering.