Balancing Statistical and Computational Trade-Offs When Extracting Selection Signal from a Large Number of DNA Sequences

• Activity Number: 64
• Type: Topic Contributed
• Date/Time: Sunday, August 3, 2014 : 4:00 PM to 5:50 PM
• Sponsor: Biometrics Section
• Author(s): Vladimir Minin*+ and Erick Matsen and Connor McCoy and Trevor Bedford
• Companies: University of Washington and Fred Hutchinson Cancer Research Center and Fred Hutchinson Cancer Research Center and Fred Hutchinson Cancer Research Center
• Keywords: empirical Bayes ; molecular evolution ; codon models ; immune response ; antibody ; imputation

Abstract

We consider a problem of estimating relative rates of synonymous and nonsynonymous substitutions (dN/dS) at individual sites in a molecular sequence alignment. This is an old problem with an almost equally long rivalry between counting and model-based methods attempting to solve this problem. Model-based approaches proceed by fitting a codon substitution model that captures heterogeneity in dN/dS across sites, providing a statistically sound way to estimate site-specific dN/dS values. Unfortunately, this strategy proves computationally prohibitive for massive data sets. By employing crude estimates of the numbers of synonymous and nonsynonymous substitutions at each site, counting approaches scale well to large data sets, but they fail to account for ancestral state reconstruction uncertainty and to provide site-specific dN/dS estimates. We propose a hybrid solution that borrows the computational strength of counting methods, but augments these methods with empirical Bayes modeling to produce a fast and reliable method capable of estimating site-specific dN/dS values in large data sets. We use our new method to study natural selection during antibody maturation.