Comprehensive Evaluation of End-Point Free Energy Methods in DNA-Ligand Interaction Predictions.

Abstract

Deoxyribonucleic acid (DNA) serves as a repository of genetic information in cells and is a critical molecular target for various antibiotics and anticancer drugs. A profound understanding of small molecule interaction with DNA is crucial for the rational design of DNA-targeted therapies. While the molecular mechanics/Poisson-Boltzmann surface area (MM/PBSA) and molecular mechanics/generalized Born surface area (MM/GBSA) approaches have been well established for predicting protein-ligand binding, their application to DNA-ligand interactions has been less explored. In this study, we systematically investigated the binding of 13 diverse small molecules to DNA, evaluating the accuracy of DNA-ligand interaction predictions across different solvation approaches, interior dielectric constants (ε_in), and molecular force fields. Our results demonstrate that MM/PBSA, using energy-minimized structures (the bsc1 force field and ε_in = 20), provides the best correlation (R_p = -0.742) with experimental binding affinities, surpassing the performance of rDock scoring functions (best R_p = -0.481). Notably, the interior dielectric constant was found to significantly impact DNA-ligand binding free energy predictions, especially for MM/PBSA. Moreover, both MM/PBSA and MM/GBSA predictions (ε_in = 16 or 20) exhibited superior performance in distinguishing native-like binding modes within the top-10 poses from decoys, compared to the molecular docking tools used in this study. However, the popular docking software PLANTS demonstrates notable efficacy in predicting the top-1 binding pose. Given the considerably higher computational cost of MM/PBSA, MM/GBSA rescoring with higher ε_in = 16 or 20 is more efficient for recognizing the native-like binding poses for DNA-ligand systems. This study presents the first detailed exploration of end-point free energy calculations in the context of DNA-ligand interactions and offers valuable insights for the application of the MM/PB(GB)SA methods in this domain.