AI-Assisted Drug Discovery
Revolutionizing pharmaceutical research through advanced computational approaches, combining structural biology, molecular dynamics simulation, and machine learning for viral protein characterization and therapeutic target identification
Our AI-assisted drug discovery research integrates advanced computational methodologies with structural biology to understand viral pathogenesis and identify therapeutic targets. We specialize in molecular dynamics simulations, high-resolution structural analysis, and machine learning approaches to characterize protein conformational dynamics and drug-target interactions at the atomic level.
Our research focuses on critical viral proteins, particularly SARS-CoV-2 spike protein dynamics and evolution. Through cryo-electron microscopy, X-ray crystallography, and computational modeling, we reveal molecular mechanisms underlying viral infectivity, host adaptation, and immune evasion strategies.
By combining structural insights with evolutionary analysis and computational predictions, we aim to accelerate the development of broad-spectrum antivirals and inform rational vaccine design strategies. Our work provides fundamental understanding of viral evolution and adaptation mechanisms that can guide pandemic preparedness and therapeutic development.
AI-Assisted Drug Discovery Framework
Comprehensive framework integrating structural biology, molecular dynamics simulation, and machine learning for viral protein characterization and therapeutic target identification
The AI-assisted drug discovery framework combines structural biology, molecular dynamics simulations, and machine learning to understand viral pathogenesis and identify therapeutic targets for pandemic preparedness.
Key Research Areas
Predicting and understanding how drug molecules interact with their biological targets using advanced ML models.
- Binding affinity prediction
- Selectivity optimization
- Off-target prediction
Developing models to predict crucial molecular properties that determine drug efficacy, safety, and pharmacokinetics.
- ADMET properties
- Toxicity assessment
- Solubility prediction
Accelerating the identification of promising drug candidates through AI-powered virtual screening of large molecular libraries.
- High-throughput screening
- Lead compound identification
- Chemical space exploration
Designing novel drug molecules with desired properties using generative AI models and reinforcement learning approaches.
- De novo drug design
- Lead optimization
- Multi-objective optimization
Methodological Approaches
We employ state-of-the-art machine learning architectures specifically designed for molecular data and drug discovery applications:
Transformer Models
SMILES and molecular sequence processing
Graph Networks
Molecular graph representation learning
Generative Models
VAEs, GANs, and flow-based models
Data Integration
- - Chemical structure information
- - Biological activity data
- - Protein structure and dynamics
- - Clinical trial outcomes
- - Literature and patent data
Model Fusion
- - Ensemble learning strategies
- - Multi-task learning frameworks
- - Transfer learning approaches
- - Active learning protocols
- - Uncertainty quantification
Applications & Success Stories
Oncology
- Kinase inhibitor optimization
- Immunotherapy target identification
- Resistance mechanism prediction
- Combination therapy design
Neurological Disorders
- Blood-brain barrier penetration
- Neurotransmitter modulation
- Protein aggregation inhibitors
- Neuroprotective agents
Our research has made significant contributions to pharmaceutical research and development:
Reduction in screening time
Improvement in hit rate
Average cost savings per project
Key Publications
Exploring the regulatory function of the N-terminal domain of SARS-CoV-2 Spike protein through molecular dynamics simulation
Li, Y., Wang, T., Zhang, J., Shao, B., Gong, H., Wang, Y., Wang, Z., & Liu, T. Y.
Innovation: Discovered the NTD "wedge" model - revealing that NTD acts as a regulatory wedge controlling RBD conformational transitions in SARS-CoV-2 spike protein
Contribution: Demonstrated that complete RBD structural transition occurs only when the neighboring NTD detaches and swings away, proposing NTD-RBD interface as a potential drug target
Impact: Provided mechanistic understanding of spike protein regulation and identified novel therapeutic targeting strategy for COVID-19 drug development
Structural insights into the SARS-CoV-2 Omicron RBD-ACE2 interaction
Lan, J., He, X., Ren, Y., Wang, Z., Zhou, H., Fan, S., Qi, C., Guo, A., Wang, L., Wang, T., & Wang, X.
Innovation: Crystal structure of Omicron RBD-ACE2 complex at 2.6 Å resolution combined with MD simulation and Markov State Model analysis to reveal dynamic binding mechanisms
Contribution: Identified four key mutations (S477N, G496S, Q498R, N501Y) enhancing ACE2 binding (~2.5-fold increase) and explained antibody escape mechanisms through structural analysis
Impact: First study combining structural and computational approaches to provide dynamic view of Omicron-ACE2 interactions, informing vaccine and therapeutic strategies
Loss of Spike N370 glycosylation as an important evolutionary event for the enhanced infectivity of SARS-CoV-2
Zhang, S., Liang, Q., He, X., Zhao, C., Ren, W., Yang, Z., Wang, Z., Ding, Q., Deng, H., Wang, T., Zhang, L., & Wang, X.
Innovation: Discovered N370 glycosylation loss through T372A mutation as key evolutionary event enabling SARS-CoV-2 host expansion from animal reservoirs to humans
Contribution: Demonstrated 50-fold infectivity increase through cryo-EM structures showing N370 glycan removal facilitates spike protein open state (37% vs 14%) essential for ACE2 binding
Impact: Revealed molecular determinants of SARS-CoV-2 enhanced infectivity and cross-species transmission, providing evolutionary framework for pandemic preparedness