HaploDMF

HaploDMF reconstructs complete viral haplotypes from third-generation long-read sequencing data using deep matrix factorization to characterize intra-host diversity of RNA viruses.

Key Features:

• Third-Generation Sequencing (TGS) support: Utilizes long-read third-generation sequencing data to enable more comprehensive haplotype reconstruction than short-read approaches.
• Deep Matrix Factorization model: Employs a deep matrix factorization model with an adapted loss function to learn latent features from aligned reads for haplotype discrimination.
• Read clustering into haplotypes: Clusters aligned reads into distinct haplotypes based on learned latent features.
• Robustness across conditions: Maintains performance across datasets with varying coverage levels, numbers of haplotypes, and sequencing error rates, including cases with mid-sequence coverage drop.
• Independence from overlap size: Maintains accuracy and reliability regardless of the overlap size between reads.
• Benchmarking and validation: Benchmarked against state-of-the-art tools using simulated and real third-generation sequencing viral datasets, demonstrating competitive performance in generating complete haplotypes.

Scientific Applications:

• Viral evolution studies: Reconstruction of complete viral haplotypes to investigate evolutionary dynamics, mutation rates, and selection pressures in RNA viruses.
• Host-microbe interaction analysis: Characterization of intra-population viral genetic diversity to inform studies of virus–host and virus–microbiome interactions.

Methodology:

HaploDMF learns latent features from aligned third-generation reads using a deep matrix factorization model with an adapted loss function and clusters reads into distinct haplotypes.