Revealing Missing Protein-Ligand Interactions Using AlphaFold Predictions.

Area of Science:

• Structural Biology
• Computational Biology
• Drug Discovery

Background:

• Protein-ligand interactions are key to molecular recognition but challenging to study in flexible regions.
• Current structural biology techniques struggle to reliably characterize interactions with intrinsically disordered proteins.
• Intrinsically disordered proteins are increasingly recognized as important drug targets due to their biological relevance and disease involvement.

Purpose of the Study:

• To evaluate AlphaFold2's (AF2) capability in predicting interactions between ligands and residues in disordered protein regions.
• To identify and characterize these previously uncharacterized interactions, termed "ghost interactions".

Main Methods:

• Utilized experimentally obtained protein structures determined after AlphaFold2's training.
• Assessed AF2's prediction accuracy for regions exhibiting order-disorder transitions.
• Analyzed predicted "ghost interactions" for residue burial and evolutionary conservation patterns.

Main Results:

• AlphaFold2 models accurately predict regions prone to order-disorder transitions.
• AF2 successfully predicts residues involved in "ghost interactions" with ligands.
• These predicted residues are often buried and display distinct evolutionary conservation compared to other flexible residues.

Conclusions:

• AlphaFold2 shows promise in predicting ligand interactions with disordered protein regions.
• This capability can enhance our understanding of molecular recognition and intrinsically disordered proteins.
• Findings support the development of novel therapeutic strategies targeting intrinsically disordered proteins.