Integrating Artificial Intelligence (AI) and Molecular Simulations for Molecular Engineering

Abstract:  

AI+X has emerged as a powerful paradigm for accelerating scientific discovery. In this seminar, I will present our efforts to integrate AI with molecular simulations to address fundamental challenges in molecular engineering. First, I will discuss our work on rethinking protein language modelsso that they treat proteins as physical molecules rather than purely as sequences. Conventional protein language models ignore the thermodynamic and structural principles governing protein behavior. To overcome this limitation, we developed S-PLM and D-PLM. S-PLM incorporates three-dimensional structural information into sequence representations, while D-PLM aligns sequence embeddings with molecular dynamics trajectories through contrastive learning. These approaches significantly improve zero-shot mutation effect prediction, protein stability estimation, and enzyme functional clustering. Second, I will illustrate how a prompt-tuned protein language model can be coupled with molecular dynamics simulations to discover novel antibiofouling peptides from microbiome libraries, even under limited and low-quality training data. Finally, extending beyond proteins, I will introduce our thermodynamics-informed machine learning framework for the discovery of deep eutectic solvents. Together, these examples illustrate how embedding physical principles into AI models can bridge data driven learning and molecular-level understanding, enabling more reliable and generalizable discovery in chemical and molecular engineering.

Bio:

Dr. Qing Shao is currently an associate professor in the Chemical and Materials Engineering department at the University of Kentucky. He got his PhD under the supervision of Dr. Shaoyi Jiang at the University of Washington and did postdoctoral work under the supervision of Dr. Carol Hall at North Carolina State University. The research of his group focuses on integrating AI with other scientific and engineering approaches to advance molecular engineering for medical and energy applications, with support from NIH,NSF, DOE, and non-profit societies. He has published >40 papers, with an h-index of 37 according to Google Scholar. He is also listed among the top 2% of most-cited researchers worldwide.