A message passing algorithm for reference-guided sequence assembly from high-throughput sequencing data

Recent development of next-generation DNA sequencing platforms has dramatically increased the efficiency of sequencing genomes or targeted regions of interest within genomes. Identification of genetic variants is an important downstream application of such platforms. In this paper, we present a novel framework for processing short reads generated by next-generation sequencing platforms, and apply it to the problem of detecting single-nucleotide polymorphisms (SNPs) in the target genome. The framework relies on a bipartite graphical model and message-passing techniques to unify the quality score recalibration and variant calling steps in the downstream data processing pipeline. This technique computes posteriori probabilities of the bases in the reconstructed sequence. Simulation results demonstrate that the proposed technique can improve the variant calling accuracy compared to broadly used alternative method.