Cancer progression is driven by genetic mutations that in turn lead to dysregulation of gene expression. With current efforts in cancer genome sequencing, much knowledge has been gained towards understanding the roles played by protein-disrupting mutations in cancer; however, major challenges remain in understanding the consequences of mutations that potentially disrupt DNA regulatory sequences and in assessing their contributions to cancer. These challenges are thus a barrier to a comprehensive understanding of cancer genetics and to realizing the full promise of precision therapies.

To bridge the gap between our abilities to ident...

Read More

Cancer progression is driven by genetic mutations that in turn lead to dysregulation of gene expression. With current efforts in cancer genome sequencing, much knowledge has been gained towards understanding the roles played by protein-disrupting mutations in cancer; however, major challenges remain in understanding the consequences of mutations that potentially disrupt DNA regulatory sequences and in assessing their contributions to cancer. These challenges are thus a barrier to a comprehensive understanding of cancer genetics and to realizing the full promise of precision therapies.

To bridge the gap between our abilities to identify genomic variations through whole genome sequencing and to interpret their causal consequences in cancer biology, I will develop deep learning models, which are a subset of newly emerging methods in artificial intelligence, that will extract the regulatory code of genomic sequences by leveraging the growing comprehensive functional genomics dataset to identify regulatory sequences that are under strong evolutionary constraints.

Specifically, building upon my past work, I will first develop methods for deciphering genomic sequence-level determinants of cell identity and differentiation programs in healthy and cancerous tissues by integrating sequence modeling approaches with single-cell data. Second, I will put major effort into identifying the evolutionary constraints that are levied on regulation by integrating evolutionary genetics with sequence models of gene regulatory properties. We will apply these methods to prioritize regulatory regions of interest in cancer genome sequences.

Together these approaches will provide both a valuable resource and a methodological framework for understanding coding and especially noncoding genome variations in cancer as well as other human diseases.

Read Less