AI is increasingly demonstrating its ability to capture and preserve the physical and chemical properties of biomolecular structures, as highlighted by recent Nobel Prizes in Physics and Chemistry. However, maximizing the potential of AI-based models in bioscience studies often depends on accurate and robust molecular representations--a challenging task due to the complexity and high dimensionality of biological data. In recent years, our lab has developed novel mathematical approaches, including multiscale modeling, differential geometry, algebraic topology, and graph theory-based models, to create low-dimensional yet powerful representations for various biological datasets. These tools have proven effective in characterizing biomolecular interactions with scalability, supporting diverse molecular representations, and maintaining robustness even with lower-quality data. Our work has consistently achieved high performance in molecular property prediction benchmarks and protein-protein interaction studies. Furthermore, our team has been recognized as a top performer in the D3R Grand Challenges, a global annual competition in computer-aided drug design.
Mathematical Biology Seminar
Friday, December 6
12:00pm MST/AZ
WXLR A111
Associate Professor
University of Tennessee, Knoxville