Our lab focuses on developing computational methods to understand the heritable component of cancer and related complex traits by leveraging large-scale data. We are interested in three broad research areas. (1) Integrating molecular phenotypes to decipher disease mechanisms. Genome-wide association studies (GWAS) can tell us where to look for genetic effects on disease, but not how these effects manifest themselves at the phenotypic level. Disentangling the underlying biological mechanisms poses the next great challenge for large-scale genetics. This is compounded by the finding that most disease-causing variants are non-coding, for which the language from genotype to phenotype is largely unknown. In parallel, molecular measurements such as gene expression and chromatin activity across individuals are being collected at an unprecedented rate. These data can inform how genetic variants influence specific cellular phenotypes and provide an avenue for us to decipher genetic mechanisms. We aim to develop statistical techniques for integrating molecular data to make sense of GWAS findings. Can we identify the disease associated genes and their regulators? Can we make concrete statements about causality? Can molecular data help us efficiently identify the specific causal mutations? Or prioritize targets for drug discovery? This work involves methods related to QTL analyses, genetic prediction, and making the most of summary-level GWAS data. (2) Understanding the relationship between heritable/germline and somatic variation in cancer. Cancer is a genetic disease with incidence driven by inherited (germline) variation in the person and evolution driven by somatic events in the tumor. However, the relationships between these classes of variation are not well understood and require understanding that spans statistical genetics, epidemiology, and clinical oncology. Identical tumor types will often respond to therapy very differently, suggesting that host genetics can play an important role in personalizing treatment. We seek to learn the interactions between germline and somatic events as well as their impact on cancer progression and response to treatment. In general, do germline variants induce specific somatic changes or modulate the impact of somatic variation on the tumor? Can we identify germline modifiers of somatic events? What role does genetic ancestry play in the evolution of the tumor? Most importantly, can we build genetic predictors to guide personalized treatment? (3) Inferring complex trait architecture at genome scale. Most common complex traits are driven by thousands (or tens of thousands) of genetic variants of small effect. This presents a challenge for the traditional approach of looking at variants one hit at a time, but also an opportunity for large-scale computation methods to extract knowledge from all variants at once. What regions of the genome are unusually important for a disease? Do features observed in specific cell-types or conditions tend to harbor trait-effecting variants, and can they inform our understanding of the trait etiology? Is the disease primarily driven by variants that disrupt coding, have subtle effects on regulation, or by as-of-yet unknown features? This work involves methods related to inference of heritability, variance component (or Gaussian Process) models, and polygenic risk prediction. Though the lab's primary focus is on statistical/quantitative genetics, we collaborate broadly with experimentalist groups to validate and inform our statistical hypotheses.
