ML-Driven Drug Discovery
Systematic study on applying machine learning methods - Elastic Net (ENET), Random Forest (RF), Gradient Boosting Machine (GBM), Multi-layer Perceptron (MLP), Gaussian Process Regressor (GPR) - to accelerate TYK2-inhibitor protein-ligand free energy calculations. Source code and dataset available are open-sourced on GitHub.
Cite: Optimizing active learning for free energy calculations, J Thompson, W P Patricks, J A Feng, N A Pabon, H Xu, B B Goldman, ... & F York, Artificial Intelligence in the Life Sciences, 2022
Small Molecule Docking
Computational simulations to reveal how chiral cysteine attached to nanoparticle selects DOPA (related to dopamine) through matching hydrogen bonds.
Cite: Mesoporous encapsulated chiral nanogold for use in enantioselective reactions, Y Zhou, H Sun, H Xu, S Matysiak, J Ren, X Qu, Angewandte Chemie
Hydrogel Network Structure
Studies the effect of protonation on chitosan hydrogel network structure formation, and develops molecular model that fits well with experimentally-measured structural and mechanical properties.
Cite: Effect of pH on chitosan hydrogel polymer network structure, H Xu, S Matysiak, Chemical Communications
Ligand-Driven Self-Assembly
Computationally simulates the effect of ligand size on shaping clusters on the cell membrane surface.
Cite: Influence of Monovalent Cation Size on Nanodomain Formation in Anionic–Zwitterionic Mixed Bilayers, SJ Ganesan, H Xu, S Matysiak (co-first author)
Membrane Peptide Folding
Develops proxy models to simulate Alzheimer's pathological peptide with order of magnitude improvement in computational efficiency.
Cite: Effect of lipid head group interactions on membrane properties and membrane-induced cationic β-hairpin folding, SJ Ganesan, H Xu, S Matysiak, (co-first author)
