Publications
Journal, conference, and workshop papers and preprints.
Zhe Wang, Zijing Liu, Chencheng Xu, Yuan Yao
IEEE International Conference on Bioinformatics and Biomedicine (BIBM), 2025
Summary
Phi-Former models compound–protein interactions across atoms and molecular motifs using hierarchical representations and pairwise pretraining. Its predictions can highlight the atoms or motifs involved in an interaction.
Maodong Li, Jiying Zhang, Zhe Wang, Bin Feng, Wenqi Zeng, Dechin Chen, Zhijun Pan, Yu Li, Zijing Liu, Yi Isaac Yang
arXiv:2504.18367, 2025
Summary
The paper introduces DD-13M, a time-resolved dataset of more than 26,000 ligand–protein dissociation processes, and UnbindingFlow, a generative model trained to produce dissociation trajectories and predict off-rates.
Shengyuan Bai, Qibin Li, Zhe Wang, Nai Zhou, Nianmin Yao
AAAI Conference on Artificial Intelligence (AAAI), 2025
Summary
Adversarial Noisy Instruction Tuning improves language models on natural-language understanding tasks by training them to recognize and handle semantic distortions in noisy instructions.
Yue Wang, Zhe Wang, Ziyuan Zhao, Zijian Li, Xun Jian, Hao Xin, Lei Chen, Jianchun Song, Zhenhong Chen, Meng Zhao
IEEE Transactions on Knowledge and Data Engineering (TKDE), 2020
Summary
This work introduces a meta-path-free way to measure similarity in heterogeneous networks, reducing the need for manually specified meta-paths while capturing structural and semantic relationships.
Zijian Li, Wenhao Zheng, Xueling Lin, Ziyuan Zhao, Zhe Wang, Yue Wang, Xun Jian, Lei Chen, Qiang Yan, Tiezheng Mao
IEEE International Conference on Data Engineering (ICDE), 2020
Summary
TransN learns representations for heterogeneous networks without requiring node labels or predefined meta-paths. It combines information from individual network views by translating embeddings between them.
Yue Wang, Zhe Wang, Ziyuan Zhao, Zijian Li, Xun Jian, Lei Chen, Jianchun Song
IEEE International Conference on Data Engineering (ICDE), 2020
Summary
HowSim extends similarity measurement to heterogeneous information networks through a decay graph, capturing structural and semantic similarity without relying on manually selected meta-paths.