Multi-Commodity Flow Methods for Quasispecies Spectrum Reconstruction Given Amplicon Reads

Keywords: Next-generation sequencing. Viral quasispecies. Network flows. RNA viruses depend on error-prone reverse-transcriptase for replication within an infected host. These errors lead to a high mutation rate which creates a diverse population of closely related variants [1]. This viral population is known as a quasispecies. As breakthroughs in next-generation sequencing have allowed for researchers to apply sequencing to new areas, studying genomes of viral quasis-pecies is now realizable. By understanding the quasispecies, more effective drugs and vaccines can be manufactured as well as cost-saving metrics for infected patients implemented [2]. Given a collection of (shotgun or amplicon) next-generation sequencing reads generated from a viral sample, the quasispecies reconstruction problem is defined as: reconstruct the quasispecies spectrum, i.e., the set of sequences and respective frequencies of the sample population. Reconstructing the quasispecies spectrum is difficult for several reasons. The actual amount of variants may be obfuscated by conserved regions in the genome that extend beyond the maximum read length. Additionally, the amount of possible assignments of reads to variants in overlapping segments grows quickly. Furthermore, we are required to rank the variants by frequency. Previous approaches have utilized min-cost flows, probabilistic methods, shortest paths, and population diversity for the quasispecies spectrum assembly problem [3–6].