Revealing Missing Protein–Ligand Interactions Using AlphaFold Predictions

Abstract

Protein–ligand interactions represent an essential step to understand the bases of molecular recognition, anintense field of research in many scientific areas. Structural biology has played a central role in unveilingprotein–ligand interactions, but current techniques are still not able to reliably describe the interactions ofligands with highly flexible regions. In this work, we explored the capacity of AlphaFold2 (AF2) to estimatethe presence of interactions between ligands and residues belonging to disordered regions. As these interactionsare missing in the crystallographic-derived structures, we called them “ghost interactions”. Using aset of protein structures experimentally obtained after AF2 was trained, we found that the obtained modelsare good predictors of regions associated with order–disorder transitions. Additionally, we found that AF2predicts residues making ghost interactions with ligands, which are mostly buried and show differential evolutionaryconservation with the rest of the residues located in the flexible region. Our findings could fuel currentareas of research that consider, given their biological relevance and their involvement in diseases,intrinsically disordered proteins as potentially valuable targets for drug development.