Friday, May 11, 2007 A Survey of SNP Haplotype Inference problem Yan Zhang

SNP haplotypes have been drawing great attention for the 
study of disease mapping and for tracing population migration 
events. Since in many situations only genotype data is available, 
computational methods have been developed to identify the 
haplotype pairs carried by an individual and to estimate 
haplotype frequencies from genotype data. There are two 
major approaches to solving the inference problem: 
combinatorial methods and statistical methods. Combinatorial 
methods often state an explicit objective function that one tries 
to optimize in order to obtain a solution to the inference problem. 
Statistical methods are usually based on an explicit model of 
haplotype evolution; the inference problem is then cast as a 
maximum-likelihood or a Bayesian inference problem. In this talk, I will introduce the distinct mathematical models 
and assumptions that support these methods. By using two 
real data sets and a set of simulated data that follows coalescent 
model, I will show my experimental evaluations on four 
representative methods. I conclude that, firstly, the informative 
biological prior is the key for this inference problem; secondly, 
future research of this problem should focus on improving the 
performance of algorithms for data sets with low LD, genotyping 
errors, missing alleles, and rare haplotypes.