Pei-Kang Tsou, Ph.D., a postdoctoral researcher in Jer-Lai Kuo’s Group, received the 2021 2nd Session Academia Sinica Postdoctoral Research Fellowship.

Dr. Tsou proposed a machine learning (ML)-assisted scheme to explore the chemical space of saccharides. The scheme aims to accelerate the search for transition states (TSs) in the dissociation channels of CID-MS for saccharides. The exploration of potential energy surfaces (PESs) and locating TSs with a huge number of conformers by quantum mechanics are challenging tasks in chemistry. ML has advanced in recent years and has been developed for PES predictions and geometry optimizations. The ML-assisted scheme includes manual and auto-search protocols to generate data points for model training, and its learning cycle allows the model to learn new sugar molecules and new reaction pathways. With the scheme, the exploration of dissociation channels for disaccharides, which is considered far more complicated than that of monosaccharides, becomes feasible.
Dr. Tsou also developed a transfer-learning scheme for a model trained with DFT methods to learn high-level quantum methods and achieve chemical accuracy. The aim of the scheme is to learn about the general configuration space from fast DFT methods and the critical minima and TS from high-level approaches with fewer data points.