Reinventing dN/dS and the study of natural selection

点击翻译按钮获取中文摘要

Natural selection on genetic variation is the ultimate cause of adaptation and the formation of new species. All studies of natural selection at the genomic level require a control group: a set of variants or substitutions that have not been subject to selection. The class of variants that, by far, are the most widely used for this purpose are synonymous changes within protein coding genes. Because these variants are interspersed with nonsynonymous variants, they are a natural control group. However, overwhelming evidence that synonymous changes are often not strictly neutral has accumulated over the past three decades. In many contexts the clear evidence for selection provides investigators with ways to identify which synonymous changes are most likely to be affected by selection and which are more likely to be strictly neutral. What is missing are ways to include this information, on classified synonymous changes (neutral and selected), within quantitative evolutionary models that enable a broad range of inferences about where, when, and on what selection operates. The proposed research is based on a new class of evolutionary models (Multiple Synonymous Substitution or MSS models) that include both selected and neutral synonymous changes in estimates of evolutionary change. These models can be fitted to homologous sequence alignments to infer relative rates of amino-acid replacement changes, and selected synonymous changes, to strictly neutral synonymous changes. These types of (dN/dS) analyses go back to the origins of evolutionary genetics, and today they remain among the most commonly used tools. Our hypothesis is that the current definition of what constitutes neutrally evolving sites is importantly wrong in that it biases the estimation of selective forces (by underestimating the rates of neutral changes), and that improved estimates without that bias will provide significantly better and more accurate understanding of the action of natural selection in a broad array of evolutionary contexts. We will develop and test a new class of substitution model that uses multiple sets of synonymous substitutions (MSS models), and we will apply them to a diverse set of taxa sampled from across life's kingdoms. Estimated MSS models, discovered using a genetic algorithm that searches over the space of all possible MSS models, will be validated using internal controls, based on neutral model predictions of substitution rate speed and constancy, and using external controls, based on predicted covariates of population genomic and functional effects. We will revise an array of widely-used comparative and population genetics techniques to take advantage of MSS models, and develop new approaches that seek to identify selection on synonymous changes. Our models will be applied to a wide taxonomic range of genomic data and made available as a part of popular software package HyPhy and web application Datamonkey.