Unsupervised approaches to decode the functional impact of genetic variation

Abstract:

The increasing availability of large-scale data is transforming our ability to study genetic regulation of cell states. However, understanding how genetic variation governs cellular function and complex diseases remains a challenge, requiring new analytical frameworks capable of integrating diverse genomic datasets to infer functional relationships. This seminar will present new unsupervised computational approaches for dissecting genetic regulation of cellular phenotypes. Our analysis of evolutionary and epigenetic conservation across human cell types has identified domains under cellular constraint that encode functional determinants of cell identity. By calculating genome-wide, single base resolution cellular constraint scores, I will demonstrate their utility in fine-mapping causal variants from genome-wide association studies, improving polygenic risk models, and predicting clinical outcomes in machine learning-based cancer survival models. These findings form the basis for development of multi-omic genome-wide unsupervised machine learning frameworks and variant-to-trait models that provide powerful approaches for functional annotation of non-coding variants and partitioning disease-associated genetic variants governing complex trait and disease sub-phenotypes. I will illustrate the versatility of these methods across various experimental applications including the study of multi-lineage differentiation from pluripotent stem cells and ongoing efforts to study population-scale data to parse the genetic basis of complex diseases. These studies illustrate new strategies to bridge the gap between genomic variation and cellular function for guiding scalable and interpretable solutions to advance our understanding of human development, disease, and therapeutic discovery.

9:30 – 10:30 | Teams | followed by refreshments in the atrium

Find out more