We work at the intersection of population genetics, statistical genetics, and machine learning. We use a combination of conceptual models, pen-and-paper analysis of mathematical models, and large-scale data analysis to understand human genetics: from how mutations impact cells all the way up to how those mutations are affected by natural selection over hundreds of thousands of years. Some of the things we are particular interested in are:

Methods for inferring and interpreting evolutionary constraint

The mutations that we see in a sample of present-day individuals are an extremely biased subset of all possible mutations: they must have survived the sieve of natural selection. That means that we can use patterns of variation to figure out what parts of the genome are constrained by natural selection and infer that those parts must be functionally important.

Unfortunately, evolution is an extremely noisy process. The frequency of a genetic variant is driven both by natural selection and random genetic drift. This makes it very difficult to tell if any single variant is being affected by selection.

To get around this, we are interested in developing machine learning models to pool information across variants by leveraging variant-level features (e.g., chromatin accessibility data or deep mutational scanning data). We are also interested in the other side of this coin: what variant-level features cause a variant to be selected against.

Relevant papers:

Causes and consequences of trait specificity

Genetic variants are often pleiotropic, meaning that they impact multiple traits. One explanation of this is the omnigenic model: the gene regulatory network within cells is sufficiently well-connected such that perturbing the expression of one gene affects the expression of many others, including genes that have direct impacts on different traits. Another explanation is that the machinery for different key cellular processes is shared across many cell types, and disrupting this machinery affects all of these cell types and the traits they influence.

In recent work we showed that trait-specific variants contribute more to heritability and are prioritized higher by association studies. Our work also hinted at a broad spectrum of specificity across genes and variants, including highly trait-specific variants acting on highly pleiotropic genes. We are interested in understanding the cellular mechanisms by which such trait specificity can arise.

Relevant paper:

Genetic impact of stabilizing selection on traits

Many traits, including risks for many diseases, appear to be under stabilizing selection: it is disadvantageous to have a trait value that is too high or too low. This is somewhat surprising: having too low of a disease risk can also be bad!

This results in fascinating dynamics at more granular biological scales. If a variant increases a trait that is subject to stabilizing selection, whether the variant is beneficial or deleterious depends on whether it’s acting in an individual that has too little or too much of the trait. This in turn depends on the effects of the other variants that that individual has and how the environment impacts the individual. Similar considerations play out across biological scales: increasing the expression of a gene expression program may be helpful or hurtful depending on how it impacts fitness-relevant traits, and whether in any given individual, that moves them closer to or further from having optimal trait values.

We are interested in better understanding how these effects play out across different biological scales, by using a combination of mathematical modeling and simulation. We want to use these insights to understand the real-world consequences of stabilizing selection: how does it impact what is discovered in association studies? How does it shape proteins and gene regulatory networks? Can we use these insights to improve genomic or protein language models?

Relevant papers:

Population genetics of climate change

Anthropogenic climate change is driving the sixth mass extinction. Beyond the loss of entire species, we are also witnessing threats to genetic diversity within species. This loss of genetic diversity will make it harder for surviving species to adapt to a rapidly changing environment.

With Moisés Expósito-Alonso we are using and building population genetics tools to understand how climate change has impacted and will impact genetic diversity. We are particularly interested in understanding how genetic diversity is affected by the interaction of geographical spatial structure in populations and the spatial structure of habitat loss.

Relevant papers: