An effective haplotype assembly algorithm based on hypergraph partitioning

Abstract

The haplotype assembly problem has been proven to be complex. Heuristic algorithms are the main methods that are used to solve the problem. These algorithms perform well when the SNP fragments are error-free, but they are less accurate when the error rate increases. The complex relationships caused by fragment errors present a major barrier to assembling accurate haplotypes. Therefore, modeling the complex relationships is the key to solve the problem. In this study, we model the haplotype assembly problem using hypergraph partitioning formulations and propose a novel method termed HGHap (Hypergraph-based Haplotype assembly method). HGHap approaches the haplotype assembly problem in two phases. In the first phase, a hypergraph is constructed in which each vertex corresponds to a fragment and vertices are multiply connected to form hyperedges. In the second phase, a hypergraph partitioning algorithm is employed to obtain two groups of fragments to construct the haplotypes. The hyperedges capture higher-order relationships among fragments that facilitate the subsequent partitioning. Our results demonstrate that the method performs better than other methods in most cases, especially in cases with a high error rate.