Research Statement: First-principle graph equivariant machine learning for molecular modeling and drug discovery
Yuanqing Wang · 2024
I design graph-based equivariant machine learning models with physics-inspired inductive biases. More specifically, I will expedite the process of:• Structure-based drug discovery with physics-based modeling (Aim 1)—fast and stable E(3)-equivariant graph models for constructing and sampling from force fields;• Ligand-based drug discovery data-driven modeling (Aim 2) with graph neural networks transcending the convolutional scheme, which will form the backbones active learning frameworks and foundation models, emphasizing data efficiency, and uncertainty quantification.