Viral quasispecies reconstruction via tensor factorization with successive read removal

Area of Science:

• Virology
• Bioinformatics
• Computational Biology

Background:

• RNA viruses evolve into viral quasispecies due to mutations and environmental adaptation.
• High-throughput sequencing (HTS) aids viral quasispecies studies, but challenges remain in strain reconstruction and spectrum estimation due to sequencing errors and read length limitations.
• Accurate inference of viral quasispecies is complicated by non-uniform strain frequencies and small genetic distances between strains.

Purpose of the Study:

• To present TenSQR, a novel algorithm for reconstructing viral quasispecies from HTS data.
• To address the challenge of analyzing viral populations with highly uneven strain frequencies and small genetic variations.
• To enable the discovery of rare viral strains and detect deletions within them.

Main Methods:

• TenSQR employs a tensor factorization framework for HTS data analysis.
• The algorithm utilizes clustering with successive data removal to infer strains from most to least abundant.
• Sequencing reads attributed to an inferred strain are removed to facilitate the discovery of rarer strains and deletions.

Main Results:

• TenSQR successfully reconstructs full-length viral strains with widely varying abundances.
• The algorithm outperforms state-of-the-art methods in datasets with 1-10% diversity.
• TenSQR effectively detects long deletions, even in rare viral strains, as demonstrated on simulated, HIV-1, and Zika virus datasets.

Conclusions:

• TenSQR provides a robust method for viral quasispecies reconstruction from HTS data.
• The algorithm's ability to identify rare strains and deletions offers significant advantages for viral evolution and drug resistance studies.
• TenSQR enhances the analysis of complex viral populations, contributing to the development of targeted antiviral strategies.