Professor Purdue University West Lafayette, Indiana
Over the past few years, we have developed unsupervised manifold embedding and manifold kernelization methods to capture electronic properties on molecular surfaces and predict molecular properties. This work is motivated by the desire to understand intermolecular interactions through electronic calculations. By deriving a kernel matrix from the electronic properties using the graph of a triangulated surface mesh through spectral analysis, we can have a reduced-dimensional representation of molecule's quantum information that is invariant to translation and rotation, making it suitable for machine learning. The manifold kernelization and learning concept has created a new research avenue for predictive and generative AI in drug discovery and development.
Learning Objectives:
Upon completion, participant will be able to appreciate a forefront in AI development for drug discovery, why it is important, how it is implemented, and what it can do
