CAREER: Algorithms for Domain-Level Analysis of Gene Family Evolution

点击翻译按钮获取中文摘要

The genome of an organism helps to determine its biology. Understanding how different genes evolve and acquire new functions is a fundamental biological problem with many computational methods developed for studying how gene families evolve and change over time in different organisms. These existing methods assume that the gene is the basic unit of evolution and that evolutionary processes such as gene duplication, gene loss, and horizontal gene transfer act on entire genes, rather than on parts of genes. It is well known that most genes consist of one or more "protein domains," well-characterized functional units that can be independently lost or gained during evolution, and that domain shuffling is one of the primary mechanisms through which genes evolve and gain new functions. Proper inference and accounting of domain-level evolutionary events is therefore crucial to understanding how genes evolve and function. The proposed research will lay the methodological and algorithmic foundations for a novel computational framework that addresses this critical problem. The new computational framework and algorithms will enable more powerful comparative genomic techniques for understanding gene function and biology, and may also contribute to improvements in human health and agriculture. The proposed research will shape future computational advances in the study of domain, gene, and genome evolution for many years to come, and will also spur the development of more comprehensive computational models in other areas of molecular evolution. The algorithms developed as part of this research will be implemented into a user-friendly software package and made freely available. The project will directly involve two graduate and up to ten undergraduate students, introduce several high-school students to computer science, bioinformatics, and research, and provide training to many high-school science teachers on the role of computer science in biology. This project will lead to the development of the first "three-tree" model of domain evolution that explicitly captures the interdependence of domain-, gene-, and species-level evolution. The proposed three-tree computational framework is based on phylogenetic reconciliation, where the goal is to find a most parsimonious joint reconciliation of the given gene trees with the species tree and of the given domain trees with the gene trees. The resulting optimization problems will be solved using various algorithmic techniques including dynamic programming, branch and bound, enumeration and sampling, and local search. The framework will decouple domain-level events from gene-level events and provide a fine-grained view of gene family and domain family evolution that is both more accurate and much easier to interpret. Specific aims include: (i) development of the three-tree computational framework and corresponding algorithms, (ii) enhancing inference accuracy by accounting for multiple optima and domain tree errors, and (iii) extension of the three-tree framework to microbial gene families by allowing for horizontal gene transfer.