Statistical Analysis of SNP Alignment From Homologous Chromosomes

In some sequencing projects, the targets contains two homologous chromosomes and our aim is to distinguish between polymorphisms and sequencing errors. Duplication is an important aspect of evolution, and paralogous regions could be mis-aligned together in one assembly. In this case our goal is to unravel duplicated regions. The difficulty lies in the fact that origins of fragments are unknown. Our method of haplotype reconstruction from SNP alignment is based on a probabilistic model. In this paper, we first report a fast algorithm to calculate the likelihood of the observed assembly matrix as a function of haplotype frequency. Its complexity is linear with respect to the number of SNP loci. Consequently, we can define overall posterior confidence score for each haplotype segment, which can help us improve reconstruction. Also this algorithm allows us to find the maximum likelihood estimate of the haplotype frequency. We present a simulation based on the parameters of ciona intestinalis.