Polygenic prediction and evolution of complex traits

University Of Pennsylvania, Philadelphia PA

Investigators

Linked publications & trials

Abstract

The genetic variation that contributes to phenotypic differences among individuals is shaped by mutation, genetic drift and natural selection. Relatively little is known about the relative contributions of drift and selection. On the one hand, in terms of genome-wide variation drift is the stronger force. On the other, this may not be true of phenotypically important variants, of which several examples are known. Ancient DNA provides the ability to directly observe evolution in action over well-defined timescales, and to identify and quantify the effects of selection. As well as explaining how and when unique aspects of our species and populations evolved, this information is important to predict disease risk across the spectrum of human ancestry. Because genetic ancestry is correlated with environment, we do not know whether ancestry- phenotype correlations are causal. The solution is that evidence of recent selection allows us to infer causality – ancient DNA provides information that is independent of observational studies of present-day populations. We propose to develop this in several ways. First, we estimate the contribution of recent (in the past 5-10 thousand years) natural selection to phenotypic variation within Britain. We focus on Britain as a proof- of-concept because of relatively simple demography, access to large samples of ancient and present-day individuals and because so much of our knowledge about the genetic basis of complex traits is derived from this population. However, clearly this only represents a small portion of human variation. Therefore, we will develop approaches to enable studies of natural selection using ancient DNA in populations with more complex demographic histories. We will apply these approaches to data from East Asia and Africa to measure the contribution of natural selection to phenotypic variation and disease risk in these regions. Taking a comparative approach, we will identify common patterns in the response to environmental change such as the independent development of agriculture in these regions. We will contrast patterns of selection in ancient and present-day populations, asking whether it has changed or intensified due to recent shifts in environment. These questions are not restricted to organismal or disease phenotypes. Molecular phenotypes like gene expression, splicing and methylation are critical to understand the functional basis for the effects of GWAS variants and are equally critical to understand the evolution of organismal phenotypes. We will test for recent selection on these phenotypes and identify biological mechanisms that underlie genomic signals of selection. Finally, we will apply these evolutionary insights to the construction of polygenic risk scores to predict disease. Because substantive differences in risk across ancestries are likely driven by selection, identifying selected variants allows us to predict those risks while removing confounding due to genetic variation that is simply correlated with environment from unselected variants. With this approach we aim to construct polygenic scores that are accurate across the spectrum of human genetic ancestry.