Estimation of genetic diversity in viral populations from next generation sequencing data with extremely deep coverage

Abstract

In this paper we propose a method and discuss its computational implementation as an integrated tool for the analysis of viral genetic diversity on data generated by high-throughput sequencing. Most methods for viral diversity estimation proposed so far are intended to take benefit of the longer reads produced by some NGS platforms in order to estimate a population of haplotypes. Our goal here is to take advantage of distinct virtues of a certain kind of NGS platform - the platform SOLiD (Life Technologies) is an example - that has not received much attention due to the short length of its reads, which renders haplotype estimation very difficult. However, this kind of platform has a very low error rate and extremely deep coverage per site and our method is designed to take advantage of these characteristics. We propose to measure the populational genetic diversity through a family of multinomial probability distributions indexed by the sites of the virus genome, each one representing the populational distribution of the diversity per site. The implementation of the method focuses on two main optimization strategies: a read mapping/alignment procedure that aims at the recovery of the maximum possible number of short-reads; the estimation of the multinomial parameters through a Bayesian approach, which, unlike simple frequency counting, allows one to take into account the prior information of the control population within the inference of a posterior experimental condition and provides a natural way to separate signal from noise, since it automatically furnishes Bayesian confidence intervals. The methods described in this paper have been implemented as an integrated tool called Tanden (Tool for Analysis of Diversity in Viral Populations).