Maohua Yang

CycDelta is a multimodal residue-graph model for predicting permeability changes in cyclic peptides by comparing within-source permeability differences, outperforming baselines on out-of-distribution and zero-shot tasks.

This study develops DMFormer, a Transformer-based graph neural network, to analyze Gα protein conformational dynamics, identifying Switch II as a critical motif for activation.

Uni-Clip improves protein-ligand interaction modeling for undruggable and druggable targets using multimodal contrastive learning, showing significant performance gains in benchmark datasets.