WgLink

WgLink reconstructs whole-genome viral haplotypes from NGS reads by synthesizing local haplotype reconstructions into global strains for accurate haplotype frequency estimation.

Key Features:

• Integration of local reconstructions: Accepts local haplotype outputs from TenSQR or aBayesQR to initialize whole-genome assembly.
• Breadth-first search stitching: Uses a breadth-first search to stitch regional haplotypes into candidate whole-genome haplotypes while tolerating specified SNV mismatches.
• L0 + L1-regularized regression: Applies L0 + L1-regularized regression that integrates variant allele frequency data with physical linkage information across variants to estimate non-zero haplotype frequencies.
• Scalability to high coverage and large genomes: Designed to operate on high sequencing coverage and large genome sizes with reduced memory consumption and CPU time compared to some existing methods.
• Validation on datasets: Demonstrated enhanced accuracy on both simulated and real sequencing datasets.

Scientific Applications:

• Whole-genome viral strain reconstruction: Reconstruction of complete viral strains from tiled local haplotypes for virology research.
• Analysis of high-coverage, large-genome sequencing data: Application to datasets with high sequencing coverage and large viral genome sizes.
• Method development and evaluation: Use in benchmarking and validating haplotype reconstruction performance on simulated and real datasets.

Methodology:

Local haplotypes are obtained using TenSQR or aBayesQR; regional haplotypes are stitched into whole-genome candidates via breadth-first search with SNV mismatch tolerance; final haplotype frequencies are estimated using L0 + L1-regularized regression that combines variant allele frequency and physical linkage information.