SG-ML-PLAP: A structure-guided machine learning-based scoring function for protein-ligand binding affinity...

Area of Science:

• Computational chemistry
• Drug discovery
• Machine learning

Background:

• Predicting protein-ligand binding affinity is crucial for drug design.
• Machine learning (ML) is increasingly used, but accurate compound ranking remains a challenge.

Purpose of the Study:

• To develop a novel ML-based scoring function (MLSF) for accurate binding affinity prediction.
• To evaluate the performance of the developed MLSF against existing methods.

Main Methods:

• Utilized extended connectivity interaction fingerprints (ECIF) from the PDBbind dataset to build ML models.
• Benchmarked performance on the Comparative Assessment of Scoring Functions (CASF) dataset and unseen complexes.
• Investigated the impact of including redocked protein-ligand complexes in the training set.

Main Results:

• The developed MLSF, SG-ML-PLAP, demonstrated improved binding affinity prediction accuracy.
• Supplementing crystal structures with redocked complexes enhanced MLSF performance.
• MLSF trained on crystal structures with ECIF and VINA features showed high accuracy for both crystal and docked complexes.

Conclusions:

• The proposed MLSF offers superior performance compared to conventional scoring functions and other MLSFs.
• SG-ML-PLAP is a valuable tool for structure-based virtual screening and identifying novel inhibitors.
• The SG-ML-PLAP webserver is freely accessible for broader research use.