Population genetics-based codon models

University Of Tennessee Knoxville, Knoxville TN

Investigators

Abstract

Mammals, viruses, bacteria, and even cancer tumors evolve. Understanding their evolutionary history can be critical to understanding their biology and potential for future change. Scientists currently use models to help infer this history, but most of these models ignore the reality of natural selection. This research proposal will develop models that allow scientists to estimate these histories while incorporating selection for certain amino acids into the models. These new models include parameters reflecting processes from population genetics, and will be used to infer the strength of selection on codon usage and amino acid sequence, sensitivity of protein function to different amino acid properties such as size and polarity, and mutation rates. These models fit empirical data far better than traditional models, and will more accurately infer evolutionary histories and processes. While the preliminary work uses amino acid sequences, the final project will use DNA sequences, allowing even greater resolution of evolutionary events. Given the importance of understanding evolutionary history and processes in fields ranging from medicine to forensics and from agriculture to basic taxonomy, the improvements in methods developed here will have great utility to both science and society. This research will create models linking mutation, drift, codon selection, and amino acid selection for use in phylogenetic inference. Rather than symmetric transition matrices derived from empirical estimates as used in the past, the models developed here will allow for twenty different transition matrices (one for each possible optimal amino acid) which allow for different rates of gain and loss between pairs of amino acids but which are generated from just a few, realistic parameters. The models developed here will be evaluated both for fit, using measures such as AIC, and adequacy, comparing them with each other and with standard models. Performance will be evaluated using multiple gene datasets from groups of organisms, ranging from a handful of genes to entire genomes, as well as from simulated genetic datasets. Preliminary analyses using a simplification of the proposed models show a dramatic improvement in model fit compared to traditional models. They also better match and predict empirical data, which is a standard test of model adequacy. Models will be incorporated into existing phylogenetics software via a hackathon including software developers. Outreach will include work with local teachers and development of a learning module for their high school classes.