Antibody-Antigen Affinity Prediction
Deep learning for predicting antibody-antigen binding affinity
Challenges
Traditional wet-lab experimentation for determining antibody-antigen binding affinity is time-consuming, expensive, and limits the speed of therapeutic antibody development and drug discovery.
Solutions
Deep learning models using Protein Language Model (PLM), Protein 3D structure prediction and graph neural networks to analyze protein structures and predict binding affinity computationally, enabling rapid screening of antibody candidates.
Key Results
- Unrelated Complex evaluation: 0.64
- Local Perturbation evaluation: 0.66
Technology
Pytorch, PyG, Graph Neural Network, Protein 3D Structure Prediction, Protein Language Model, Biopython