Artificial Intelligence Tools for Antibody Engineering

In a blind prediction challenge in 2020, AlphaFold2 calculated highly accurate three-dimensional structures of hundreds of proteins from their sequences, thus “solving…one of the biggest problems in biology”.1 This achievement and others based on artificial intelligence (AI) algorithms have unlocked incredible possibilities for biomolecular engineering. In this talk, I will share advances from my lab in antibody engineering and protein-protein docking based on AI.

Neural network models outperform physical models for antibody structure prediction, and generative language models offer multiple promising routes for design of antibody therapeutics. Docking methods reveal biological mechanisms and allow for screening. I will use the docking case to show how AI methods differ from physics-based approaches, suggesting ways to benefit from their combination.

1 Scientific American, October 2022