LeScore: a scoring function incorporating hydrogen bonding penalty for protein-ligand docking

Area of Science:

• Computational chemistry
• Structural biology
• Drug discovery

Background:

• Molecular docking is crucial for structure-based virtual screening, relying heavily on accurate scoring functions.
• Existing scoring functions often fail to adequately model the disruption of solvent hydrogen bonds, impacting binding energy prediction accuracy.
• This limitation hinders precise virtual screening, especially when polar interactions are critical.

Purpose of the Study:

• To introduce LeScore, a novel scoring function designed to improve molecular docking accuracy.
• To specifically address the inadequate accounting of solvent hydrogen bond breakage in current scoring functions.
• To enhance the prediction of binding energy by penalizing unfavorable polar interactions.

Main Methods:

• LeScore was developed as a linear combination of descriptors, including van der Waals, hydrogen bond energy, ligand strain, and a new hydrogen bonding penalty (HBP).
• The function was optimized using multiple linear regression (MLR) on the PDBbind 2019 dataset, evaluating 12 descriptor combinations.
• Performance was assessed using Pearson and Spearman correlation coefficients and validated on the Directory of Useful Decoys: Enhanced (DUD-E) dataset.

Main Results:

• LeScore achieved a Pearson correlation coefficient (rp) of 0.53 in the training set and 0.52 in the testing set.
• The scoring function demonstrated improved screening capability on a DUD-E subset, achieving an AUC of 0.71 for specific targets.
• LeScore outperformed models lacking the HBP, enhancing the ranking and classification of active compounds.

Conclusions:

• LeScore offers a robust advancement for virtual screening by incorporating a specific penalty for broken hydrogen bonds in aqueous environments.
• This novel approach improves the accuracy of binding energy predictions, particularly for ligands where hydrogen bonding is essential.
• LeScore provides a valuable tool for more effective structure-based drug discovery and lead optimization.