|
Activity Number:
|
594
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Thursday, August 6, 2009 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #305490 |
|
Title:
|
Modeling Indel Hot Spots When Inferring Alignments
|
|
Author(s):
|
Benjamin D. Redelings*+
|
|
Companies:
|
North Carolina State University
|
|
Address:
|
Bioinformatics Research Center, Raleigh, NC, 27695,
|
|
Keywords:
|
Bayesian ; MCMC ; phylogeny ; alignment ; indel ; transducer
|
|
Abstract:
|
We improve MCMC estimates of phylogenies and multiple sequence alignments by extending the probabilistic model of insertions and deletions to allow spatial variation in the insertion-deletion rate. This extension should improve model fit and improve parameter estimates because the local indel rate affects both the number of indels that are inferred and the strength of evidence for common ancestry for each shared indel. We annotate alignment columns with rate labels, and present an algorithm for sampling from the posterior distribution of rate labels in linear time given a fixed alignment. This algorithm assumes that the insertion-deletion process can be represented using probabilistic transducers, and we present such a model. We initially focus on models in which there are only two rates, and seek to estimate model parameters given a fixed alignment.
|
