|
Activity Number:
|
125
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 4, 2003 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Bayesian Stat. Sciences
|
| Abstract - #300823 |
|
Title:
|
Bayesian Methods for Phylogeny Independent Detection of Positively Selected Amino Acid Sites
|
|
Author(s):
|
Daniel Merl*+ and Raquel Prado and Ananias Escalante
|
|
Companies:
|
University of California, Santa Cruz and Uuniversity of California, Santa Cruz and Centers for Disease Control and Prevention
|
|
Address:
|
1156 High St., Santa Cruz, CA, 95064,
|
|
Keywords:
|
Markov chain Monte Carlo ; positive selection ; sequence evolution ; Bayesian inference
|
|
Abstract:
|
A positively selected amino acid site is one for which evolutionary forces encourage diversification. The identification of such sites is of biomedical importance: diversifying sites cannot act as reliable receptors for location specific drugs. The two current methods of detection, Nielsen and Yang's maximum likelihood method (ML) and Huelsenbeck and Ronquist's Bayesian approach, use similar stochastic models of sequence evolution but differ in their assumptions regarding phylogenetic structure. ML assumes a fixed phylogeny, which may be preferable for small numbers of distantly related sequences but is generally inappropriate for large numbers of closely related sequences. The Bayesian method accounts for phylogenetic uncertainty by using MCMC to integrate over tree structures, but often involves very long convergence times for large data sets. We introduce a new Bayesian method for detecting positive selection that eliminates the need for explicit assumptions about phylogenies and offers reduced time to convergence. We compare each method's results using data of varying degrees of speciation, including a new alignment of the AMA-1 malarial antigen.
|
