Machine learning to predict de novo protein-protein interactions

Area of Science:

• Structural biology
• Computational biology
• Machine learning

Background:

• Machine learning (ML) has significantly improved the prediction of protein-protein interactions (PPIs).
• Predicting novel PPIs, especially those not observed in nature (de novo), remains a challenge.

Purpose of the Study:

• To review recent advancements in computational methods for predicting protein-protein interactions (PPIs).
• To highlight novel machine learning approaches for de novo PPI prediction.
• To explore the biotechnological applications of these predictive capabilities.

Main Methods:

• Review of recent literature on machine learning algorithms for PPI prediction.
• Focus on methods utilizing co-folding, atomic graphs, and molecular surface learning.
• Discussion of de novo prediction strategies, including those induced by small molecules.

Main Results:

• Novel ML algorithms, such as those based on co-folding and atomic graphs, show promise for PPI prediction.
• Methods learning from molecular surfaces can predict previously unknown interactions, including small molecule-induced ones.
• Emerging applications include predicting antibody-antigen complexes and molecular glue-induced PPIs.

Conclusions:

• Computational prediction of PPIs, particularly de novo prediction, is rapidly advancing.
• These advancements hold significant potential for drug discovery and protein engineering.
• Future research directions include further development of ML models and exploration of new biotechnological applications.