Inferring Piecewise Ancestral History from Haploid Sequences

There has been considerable recent interest in the use of haplotype structure to aid in the design and analysis of case-control association studies searching for genetic predictors of human disease. The use of haplotype structure is based on the premise that genetic variations that are physically close on the genome will often be predictive of one another due to their frequent descent intact through recent evolution. Understanding these correlations between sites should make it possible to minimize the amount of redundant information gathered through assays or examined in association tests, improving the power and reducing the cost of the studies. In this work, we evaluate the potential value of haplotype structure in this context by applying it to two key sub-problems: inferring hidden polymorphic sites in partial haploid sequences and choosing subsets of variants that optimally capture the information content of the full set of sequences. We develop methods for these approaches based on a prior method we developed for predicting piece-wise shared ancestry of haploid sequences. We apply these methods to a case study of two genetic regions with very different levels of sequence diversity. We conclude that haplotype correlations do have considerable potential for these problems, but that the degree to which they are useful will be strongly dependent on the population sizes available and the specifics of the genetic regions examined.