5 Feb 10:15 Pasi Rastas: Context Tree Weighting and Its Application to Haplotype Inference

HIIT seminar, Friday Feb 5, 10:15 a.m. (coffee from 10), Exactum B222

Dr. Pasi Rastas
Helsinki Institute for Information Technology HIIT
Department of Computer Science
University of Helsinki

Context Tree Weighting and Its Application to Haplotype Inference

The context tree weighting (CTW) is an efficient method to compute weighted average over all tree models (variable order Markov chains) for a given data. The primary use of CTW is data compression, as it can be combined with arithmetic coding. CTW can be seen as a Bayesian  method.

Due to the general interest among colleagues, I will present the context tree weighting method for data compression and discuss its theoretical properties.

I will also explain our newest solution named BACH (Bayesian context-based haplotyping) for haplotype inference. Haplotype inference is a computational problem (related to missing data imputation) where the goal is to estimate population haplotypes from a sample of genotypes as accurately as possible. The studied haplotypes and genotypes consist of single nucleotide polymorphisms (mutations) in the DNA of sampled individuals from different human populations.

BACH uses CTW method to efficiently evaluate its posterior probability for a given haplotype solution. BACH achieves good accuracy by searching for haplotypes with high posterior probability. 

Last updated on 3 Feb 2010 by Visa Noronen - Page created on 5 Feb 2010 by Visa Noronen