Accurate Computational Detection of Influenza Reassortments

点击翻译按钮获取中文摘要

DESCRIPTION (provided by applicant): The influenza virus is endemic in the human population and causes significant annual morbidity and mortality. One of the primary mechanisms for generating such antigenically novel influenza strains is re- assortment, in which viral segments from two distinct strains combine. Reassortment occurs frequently among human and avian isolates and is an important feature of the evolution of virus. The recent explosion in available influenza sequence data has made it a pressing need to be able to computationally detect reassortments quickly and accurately. By doing so, we will be able identify new, potentially harmful strains quickly. We will also gain a better understanding of how reassortment occurs and why certain reassortments are more evolutionarily successful than others. We propose to (Aim 1) validate and improve a new computational approach for the accurate detection of influenza reassortments. The method takes into account uncertainty in the estimated evolutionary histories of the influenza segments by comparing two distributions of phylogenetic trees, rather than a pair of possibly unreliable or uninformative consensus trees. The proposed method permits the assignment of a confidence score to each reassortment event, something that is not possible with other approaches. We propose to validate the method on collections of human and avian genomes and also on extensive simulated data. In order to further improve the methods accuracy, we propose several extensions based on novel statistical methods that assess the changes in evolutionary distances between isolates. One outcome will be a stand-alone software package. We also propose to (Aim 2) computationally construct a large catalog of reassortments involving the sequenced isolates and to use this catalog to estimate the frequency and characteristics of reassortments. In particular, we will look for sequence mutations that tend to occur contemporaneously with reassortments. By more accurately detecting these reassortment events, we will gain a better understanding of influenza evolution. This will help plan vaccination strategies and design effective surveillance protocols. PUBLIC HEALTH RELEVANCE: We propose to study new computational methods for predicting reassortments, a key event in the evolution of influenza virus, an important human pathogen. By more accurately detecting these reassortment events, we will gain a better understanding of influenza evolution. This will help plan vaccination strategies and design effective surveillance protocols.