Iain Mathieson

Ancient DNA can directly reveal the contribution of natural selection to human genomic variation. However, while the analysis of ancient DNA has been successful at identifying genomic signals of selection, inferring the phenotypic consequences of that selection has been more difficult. Most trait-associated variants are noncoding, so we expect that a large proportion of the phenotypic effects of selection will also act through noncoding variation. Since we cannot measure gene expression directly in ancient individuals, we used an approach (Joint-Tissue Imputation [JTI]) developed to predict gene expression from genotype data. We tested for changes in the predicted expression of 17,384 protein coding genes over a time transect of 4,500 years using 91 present-day and 616 ancient individuals from Britain. We identified 28 genes at seven genomic loci with significant (false discovery rate [FDR] < 0.05) changes in predicted expression levels in this time period. We compared the results from our transcriptome-wide scan to a genome-wide scan based on estimating per-single nucleotide polymorphism (SNP) selection coefficients from time series data. At five previously identified loci, our approach allowed us to highlight small numbers of genes with evidence for significant shifts in expression from peaks that in some cases span tens of genes. At two novel loci (SLC44A5 and NUP85), we identify selection on gene expression not captured by scans based on genomic signatures of selection. Finally, we show how classical selection statistics (iHS and SDS) can be combined with JTI models to incorporate functional information into scans that use present-day data alone. These results demonstrate the potential of this type of information to explore both the causes and consequences of natural selection.

Understanding natural selection and other forms of non-neutrality is a major focus for the use of machine learning in population genetics. Existing methods rely on computationally intensive simulated training data. Unlike efficient neutral coalescent simulations for demographic inference, realistic simulations of selection typically require slow forward simulations. Because there are many possible modes of selection, a high dimensional parameter space must be explored, with no guarantee that the simulated models are close to the real processes. Finally, it is difficult to interpret trained neural networks, leading to a lack of understanding about what features contribute to classification. Here we develop a new approach to detect selection and other local evolutionary processes that requires relatively few selection simulations during training. We build upon a generative adversarial network trained to simulate realistic neutral data. This consists of a generator (fitted demographic model), and a discriminator (convolutional neural network) that predicts whether a genomic region is real or fake. As the generator can only generate data under neutral demographic processes, regions of real data that the discriminator recognizes as having a high probability of being “real” do not fit the neutral demographic model and are therefore candidates for targets of selection. To incentivize identification of a specific mode of selection, we fine-tune the discriminator with a small number of custom non-neutral simulations. We show that this approach has high power to detect various forms of selection in simulations, and that it finds regions under positive selection identified by state-of-the-art population genetic methods in three human populations. Finally, we show how to interpret the trained networks by clustering hidden units of the discriminator based on their correlation patterns with known summary statistics.