AI4Science Research Projects

❤️ Welcome to our webpage showcasing a collection of projects in the field of AI4Science. This platform highlights the collaborative efforts of a group of individuals/friends from diverse backgrounds united by a common interest in AI4Science research.

Updates

  • 2024-03: 🌟The first version of the webpage is online.

Introduction

A curated selection of AI4science projects in various domains. Each project item provides an short overview into the objectives, methodologies, and outcomes of our research. We aim to share and foster a community of learners interested in the intersection of AI4science. Explore our projects, and we hope they inspire you to delve deeper into the fascinating world of AI4Science.

Projets are in the following areas:

😁If you would like to join to prompt your research, please contact any of the members. We welcome all friends!

Research Works

The following badges are used for according purpose:

Biology

BIB-2022
SPRoBERTa: Protein Embedding Learning with Local Fragment Modeling

Lijun Wu#, Chengcan Yin, Jinhua Zhu, Zhen Wu, Liang He,Yingce Xia, Shufang Xie, Tao Qin, Tie-Yan Liu


SPRoBERTa is a protein embedding learning model that leverages local fragment modeling to capture the local structure of proteins.
Arxiv-2021
Pre-training Co-evolutionary Protein Representation via A Pairwise Masked Language Model

Liang He, Shizhuo Zhang, Lijun Wu, Huanhuan Xia, Fusong Ju, He Zhang, Siyuan Liu, Yingce Xia, Jianwei Zhu, Pan Deng, Bin Shao, Tao Qin, Tie-Yan Liu


PMLM is a pairwise masked language model that pre-trains protein representations by co-evolving the protein sequence and structure.
NeurIPS-2021
Co-evolution Transformer for Protein Contact Prediction

He Zhang, Fusong Ju, Jianwei Zhu, Liang He, Bin Shao#, Nanning Zheng#, Tie-Yan Liu


Co-evolution Transformer is a novel Transformer architecture for extracting co-evolution information from multiple sequence alignments (MSAs), which can improve the performance of protein contact prediction.

Drug Discovery

BIB-2023
SMT-DTA: Breaking the barriers of data scarcity in drug–target affinity prediction

Qizhi Pei, Lijun Wu#, Jinhua Zhu, Yingce Xia, Shufang Xie, Tao Qin, Haiguang Liu, Tie-Yan Liu, Rui Yan#


SMT-DTA is a novel drug–target affinity prediction model that breaks the barriers of data scarcity by leveraging the semi-supervised multi-task training strategy.
NeurIPS-2023
FABind: Fast and Accurate Protein-Ligand Binding

Qizhi Pei, Kaiyuan Gao, Lijun Wu#, Jinhua Zhu, Yingce Xia, Shufang Xie, Tao Qin, Kun He, Tie-Yan Liu, Rui Yan#


FABind is a fast and accurate protein-ligand binding model that achieves state-of-the-art performance by leveraging a novel FABind layer design and several rational strategies.
Arxiv-2024
FABind+: Enhancing Molecular Docking through Improved Pocket Prediction and Pose Generation

Kaiyuan Gao, Qizhi Pei, Jinhua Zhu, Tao Qin, Kun He, Lijun Wu#


FABind+ is an improved version of FABind, which not only achieves better performance by designing more rational pocket prediction method but also extend to be a sampling docking model by leveraging a simple and effective sampling strategy.
ICML-2023
Retrosynthetic Planning with Dual Value Networks

Guoqing Liu, Di Xue, Shufang Xie, Yingce Xia, Austin Tripp, Krzysztof Maziarz, Marwin Segler, Tao Qin#, Zongzhang Zhang, Tie-Yan Liu


PDVN is a retrosynthetic planning model that leverages dual value networks to generate high-quality and diverse retrosynthetic routes.
KDD-2022
RetroGraph: Retrosynthetic Planning with Graph Search

Shufang Xie, Rui Yan#, Peng Han#, Yingce Xia, Lijun Wu, Chenjuan Guo, Bin Yang, Tao Qin


RetroGraph is a retrosynthetic planning model that leverages graph search to generate high-quality and diverse retrosynthetic routes.
ICLR-2023
O-GNN: incorporating ring priors into molecular modeling

Jinhua Zhu, Kehan Wu, Bohan Wang, Yingce Xia#, Shufang Xie, Qi Meng, Lijun Wu, Tao Qin, Wengang Zhou, Houqiang Li, Tie-Yan Liu


O-GNN is a novel molecular modeling model that incorporates ring priors into the graph neural network to improve the performance of molecular property prediction.
ICLR-2023
MiCaM: De Novo Molecular Generation via Connection-aware Motif Mining

Zijie Geng, Shufang Xie#, Yingce Xia, Lijun Wu, Tao Qin, Jie Wang#, Yongdong Zhang, Feng Wu, Tie-Yan Liu


MiCaM is a novel de novo molecular generation model that leverages connection-aware motif mining to generate diverse and high-quality molecules.
TMLR-2022
Direct molecular conformation generation

Jinhua Zhu, Yingce Xia#, Chang Liu#, Lijun Wu, Shufang Xie, Tong Wang, Yusong Wang, Wengang Zhou, Tao Qin, Houqiang Li, Tie-Yan Liu


DMCG is a direct molecular conformation generation model that generates molecular conformations directly from SMILES strings.
KDD-2023
Pre-training Antibody Language Models for Antigen-Specific Computational Antibody Design

Kaiyuan Gao, Lijun Wu#, Jinhua Zhu, Tianbo Peng, Yingce Xia, Liang He, Shufang Xie, Tao Qin, Haiguang Liu, Kun He, Tie-Yan Liu


AbGNN is a pre-training antibody language model that learns antigen-specific antibody representations for computational antibody design.
KDD-2023
Dual-view Molecular Pre-training

Jinhua Zhu, Yingce Xia#, Tao Qin, Wengang Zhou, Houqiang Li, Tie-Yan Liu


DVMP is a dual-view molecular pre-training model that learns molecular representations by jointly pre-training on graph and sequence molecular data.
KDD-2022
Unified 2D and 3D Pre-Training of Molecular Representations

Jinhua Zhu, Yingce Xia#, Lijun Wu, Shufang Xie, Tao Qin, Wengang Zhou, Houqiang Li, Tie-Yan Liu


A unified molecule pre-training model that learns molecular representations by jointly pre-training on 2D and 3D molecular data.

LLM4Science

ACL-2024 (Findings)
BioT5+: Towards Generalized Biological Understanding with IUPAC Integration and Multi-task Tuning

Qizhi Pei, Lijun Wu#, Kaiyuan Gao, Xiaozhuan Liang, Yin Fang, Jinhua Zhu, Shufang Xie, Tao Qin, Rui Yan#


BioT5+ is a pre-trained language model that integrates IUPAC nomenclature and multi-task tuning to achieve generalized biological understanding, which is an extension of BioT5.
EMNLP-2023
BioT5: Enriching Cross-modal Integration in Biology with Chemical Knowledge and Natural Language Associations

Qizhi Pei, Wei Zhang, Jinhua Zhu, Kehan Wu, Kaiyuan Gao, Lijun Wu#, Yingce Xia#, Rui Yan#


BioT5 is a pre-trained language model that enriches cross-modal integration in biology with chemical knowledge and natural language associations.
ACL-2023
MolXPT: Wrapping Molecules with Text for Generative Pre-training

Zequn Liu, Wei Zhang, Yingce Xia#, Lijun Wu, Shufang Xie, Tao Qin, Ming Zhang#, Tie-Yan Liu


MolXPT is a generative pre-training model that wraps molecules with text to learn molecular representations and generate molecular structures.
ICLR-2024
Mol-Instructions: A Large-Scale Biomolecular Instruction Dataset for Large Language Models

Yin Fang, Xiaozhuan Liang, Ningyu Zhang#, Kangwei Liu, Rui Huang, Zhuo Chen, Xiaohui Fan, Huajun Chen#


Mol-Instructions is a large-scale biomolecular instruction dataset that provides detailed instructions for various biomolecular tasks, which can be used to pre-train large biomolecule language models.
Arxiv-2024
Leveraging Biomolecule and Natural Language through Multi-Modal Learning: A Survey

Qizhi Pei, Lijun Wu#, Kaiyuan Gao, Jinhua Zhu, Yue Wang, Zun Wang, Tao Qin, Rui Yan#


This survey paper reviews the recent advances in leveraging biomolecule and natural language through multi-modal learning for various biological tasks.
NeurIPS-2023
De novo Drug Design using Reinforcement Learning with Multiple GPT Agents

Xiuyuan Hu, Guoqing Liu#, Yang Zhao, Hao Zhang#


MolRL-MGPT is a de novo drug design model that leverages reinforcement learning with multiple GPT agents to generate diverse and high-quality molecules.
Arxiv-2023
The impact of large language models on scientific discovery: a preliminary study using gpt-4

Microsoft Research AI4Science, Microsoft Azure Quantum


This paper studies the impact of large language models on scientific discovery using GPT-4, and provides a preliminary analysis of the potential of LLMs in scientific research.

Computation Chemistry

NeurIPS-2023
Efficiently incorporating quintuple interactions into geometric deep learning force fields

Zun Wang, Guoqing Liu, Yichi Zhou, Tong Wang, Bin Shao


QuinNet is a geometric deep learning force field that efficiently incorporates quintuple interactions to improve the accuracy of molecular dynamics simulations.
Nature Communication-2024
Enhancing geometric representations for molecules with equivariant vector-scalar interactive message passing

Yusong Wang, Tong Wang#, Shaoning Li, Xinheng He, Mingyu Li, Zun Wang, Nanning Zheng, Bin Shao, Tie-Yan Liu


VisNet is a geometric deep learning model that enhances geometric representations for molecules with equivariant vector-scalar interactive message passing, which improves the performance of molecular dynamics simulations.
Nature Computational Science-2024
Overcoming the barrier of orbital-free density functional theory for molecular systems using deep learning

He Zhang, Siyuan Liu, Jiacheng Liu, Chang Liu#, Shuxin Zheng#, Ziheng Liu, Tong Wang, Nanning Zheng, Bin Shao#


M-OFDFT is a deep learning implementation of orbital-free density functional theory (OFDFT) that achieves DFT-level accuracy on molecular systems with lower cost complexity, and can extrapolate to much larger molecules than those seen during training.
ICML-2024
Self-consistency training for Hamiltonian prediction

He Zhang, Chang Liu, Zun Wang, Xinran Wei, Siyuan Liu, Nanning Zheng, Bin Shao, Tie-Yan Liu


The paper reveals that Hamiltonian prediction possesses a self-consistency principle, and hence proposes an extact variational training method without labeled data.
ICLR-2024
Long-Short-Range Message-Passing: A Physics-Informed Framework to Capture Non-Local Interaction for Scalable Molecular Dynamics Simulation

Yunyang Li, Yusong Wang, Lin Huang, Han Yang, Xinran Wei, Jia Zhang, Tong Wang, Zun Wang, Bin Shao, Tie-Yan Liu


We proposed a long-range short-range message-passing framework to capture non-local interactions and demonstrate the state-of-the-art results with up to 40% error reduction for molecules in MD22 and Chignolin datasets.

PDE

ICLR-2024
Better Neural PDE Solvers Through Data-Free Mesh Movers

Peiyan Hu, Yue Wang, Zhi-Ming Ma


DMM is a data-free mesh mover that can improve the performance of neural PDE solvers by learning the mesh movement strategy, which can be embedded into the neural PDE solver through proper architectural design, called MM-PDE.
ICML-2023
NeuralStagger: Accelerating Physics-constrained Neural PDE Solver with Spatial-temporal Decomposition

Xinquan Huang, Wenlei Shi, Qi Meng, Yue Wang, Xiaotian Gao, Jia Zhang, Tie-Yan Liu


NeuralStagger is a physics-constrained neural PDE solver that accelerates the convergence of PDE solvers by spatial-temporal decomposition.

Awesome-Repos

Members

(ordered by last name)

  • Kaiyuan Gao, Huazhong University of Science and Technology
  • Xiaozhuan Liang, Zhejiang Univeristy
  • Chang Liu, Microsoft Research AI4Science
  • Guoqing Liu, Microsoft Research AI4Science
  • Qizhi Pei, Renmin University of China
  • Lijun Wu, Microsoft Research AI4Science
  • Yue Wang, Microsoft Research AI4Science
  • Zun Wang, Microsoft Research AI4Science
  • Yingce Xia, Microsoft Research AI4Science
  • Shufang Xie, Microsoft Research AI4Science
  • He Zhang, Xi'an Jiaotong Univeristy
  • Jinhua Zhu, University of Science and Technology of China