Evaluation of 11 scoring functions performance on matrix metalloproteinases

Area of Science:

• Biochemistry and Molecular Biology
• Computational Chemistry
• Pharmacology

Background:

• Matrix metalloproteinases (MMPs) are crucial in physiological and pathological processes, including inflammation and cancer.
• Developing effective MMP inhibitors requires accurate computational methods for predicting ligand-enzyme interactions.
• Evaluating the performance of various scoring functions is essential for optimizing drug discovery pipelines.

Purpose of the Study:

• To assess the performance of eleven scoring functions in predicting binding affinities of MMP inhibitors.
• To evaluate the ability of these scoring functions in re-ranking virtual screening results for MMP-12.
• To develop a Principal Component Analysis (PCA) model using top-performing functions to improve enrichment.

Main Methods:

• Correlation analysis between scoring function predictions and experimental binding affinities for 3D ligand-MMP complexes.
• Virtual screening enrichment analysis using Enrichment Factor and Receiver Operating Characteristic (ROC) curves for MMP-12.
• Development of a PCA model integrating the best scoring functions (ChemScore, AutoDock, DSX).

Main Results:

• F-Score, DSX, and ChemScore showed the best overall performance in predicting MMP inhibitor binding affinities.
• ChemScore, AutoDock, and DSX exhibited the highest discriminative power in virtual screening against MMP-12.
• A PCA model combining ChemScore, AutoDock, and DSX significantly improved overall virtual screening enrichment.

Conclusions:

• ChemScore, DSX, and F-Score are recommended for predicting MMP inhibitor binding affinities.
• ChemScore, AutoDock, and DSX are effective for virtual screening of MMP inhibitors.
• A PCA-based consensus scoring approach can enhance the efficiency of MMP inhibitor discovery.