Venue: 1210 Chemistry & Willard H Dow Laboratory
Bio: Aaron Frank is originally from Grenada, a small island in the Caribbean. After moving to the US in 2001, Aaron received his BA in chemistry from Brooklyn College in 2006, where he carried out research in the groups of Professors Charlene Forest, Shaneen Singh, and Alexander Greer. He then moved to Michigan to attend graduate school at the University of Michigan and then, with his Ph.D advisor Professor Ioan Andricioaei, moved to UC Irvine in 2008. Aaron received his Ph.D in chemistry in 2011. Following a 2 year stint at Nymirum Inc. — a small biotech company in Ann Arbor founded by a close collaborator, Professor Hashimi Al-Hashimi — he returned to the University of Michigan as a Presidential Postdoctoral Fellow where he was mentored by Professor Charles L. Brooks, III. Aaron is now an Assistant Professor at the University of Michigan in the Chemistry Department and the Biophysics Department.
In this talk, I will describe examples of how my research group uses data science tools to tackle research problems that fall at the interface between Biology, Chemistry, and Physics. First, I will describe ongoing research focused on mapping the structure-landscape of functional ribonucleic acids (or RNAs). In this project, we combined machine learning and secondary structure modeling tools to predict the structure of RNAs conditioned on available NMR chemical shift data. This method now enables us to model individual conformational states, including previously invisible states of an RNA, based on its sequence and available chemical shift data. Second, I will describe ongoing research centered around decoding structure-kinetic relationships (SKRs) in sparse datasets. There is now immense interest in developing drugs that exhibit elevated residence times on their target. In this project, we used machine learning to encapsulate SKRs for CDK2, a prominent cancer target, from a dataset containing only fourteen (14) samples. I will describe our efforts to build and test CDK2-specific SKR models that take as input, the atomic structure of receptor-ligand complexes and output estimates of their residence times. Additionally, I will describe proof-of-concept studies that demonstrate the utility of our CDK2-specific SKR models as tools to help efficiently explore chemical space in search of novel chemical scaffolds that are enriched with high-residence time and potent inhibitors of CDK2.