Leonardo De Maria

The paper develops a machine learning framework to predict MHC class II binding for canonical and modified peptides by incorporating chemical features, improving immunogenicity risk assessment for non-canonical peptide drugs.

This study compares the substrate scope of a modern and ancestral family-1 glycosidase, finding the ancestral enzyme more efficient with large aglycon-containing flavonoid glycosides.

The paper benchmarks free energy perturbation calculations for predicting mutation effects on protein-protein binding affinity and introduces an automated method to improve accuracy.

The paper combines deep learning and mechanistic modeling to predict protein conformational ensembles and compares them to experimental MD structures, showing correlations and highlighting areas for improvement in folding and dynamics prediction.