HaploDMF: viral haplotype reconstruction from long reads via deep matrix factorization

Area of Science:

• Virology
• Bioinformatics
• Genomics

Background:

• RNA viruses evolve rapidly due to limited proofreading, generating diverse genomes (haplotypes).
• Characterizing viral haplotypes is crucial for understanding viral evolution and host interactions.
• Next-generation sequencing's short reads hinder complete haplotype reconstruction.

Purpose of the Study:

• To introduce HaploDMF, a novel tool for complete viral haplotype reconstruction.
• To leverage third-generation sequencing (TGS) data for improved haplotype resolution.

Main Methods:

• HaploDMF employs a deep matrix factorization model with a custom loss function.
• The model learns latent features from aligned sequencing reads to cluster haplotypes.
• It is designed for robustness against varying sequencing coverage, haplotype numbers, and error rates.

Main Results:

• HaploDMF successfully reconstructs complete viral haplotypes using TGS data.
• The tool demonstrates robust performance across diverse sequencing conditions and viral datasets.
• It outperforms existing methods, particularly in scenarios with low or uneven sequencing coverage.

Conclusions:

• HaploDMF offers a significant advancement in viral population genomics.
• The tool enhances the ability to study viral evolution and adaptation.
• HaploDMF provides a reliable method for complete haplotype reconstruction from TGS data.