Research line

We are using AI-based methods like AlphaFold to improve structural modeling of protein-protein complexes. Our current approach consists in generating hundreds of structural models with AlphaFold, which are scored by pyDock energy-based function. We are exploring the use of additional descriptors to build new ML-based models for better scoring functions.

Validation of methodology is essential in order to define the predictive capabilities in realistic applications. For that, we are participating in community-wide assessment initiatives like CAPRI or CASP.