Modern machine learning has been found to be an essential tool of the process of scientific discovery. Specifically, graph machine learning methods have achieved huge success in modeling structures such as proteins, molecules, and materials. This project will bring applicants into the interface of computer science and natural science. Participants will survey the recent progress in machine learning modeling of physical, chemical, and biological systems. Participants will be able to join ongoing projects or initiate new research projects to handle real-world problems. Potential research directions include (1) Graph ML for chemical reaction modeling; (2) Graph ML for SMLM; (3) Graph ML for plasma system inversion; and (4) Graph ML for multi-body particle system.