Prediction of Binding Free Energy of Protein–Ligand Complexes with a Hybrid Molecular Mechanics/Generalized Born Surface Area and Machine Learning Method

Abstract

Accurate prediction of protein–ligand binding free energies is important in enzyme engineering and drug discovery. The molecular mechanics/generalized Born surface area (MM/GBSA) approach is widely used to estimate ligand-binding affinities, but its performance heavily relies on the accuracy of its energy components. A hybrid strategy combining MM/GBSA and machine learning (ML) has been developed to predict the binding free energies of protein–ligand systems. Based on the MM/GBSA energy terms and several features associated with protein–ligand interactions, our ML-based scoring function, GXLE, shows much better performance than MM/GBSA without entropy. In particular, the good transferability of the GXLE model is highlighted by its good performance in ranking power for prediction of the binding affinity of different ligands for either the docked structures or crystal structures. The GXLE scoring function and its code are freely available and can be used to correct the binding free energies computed by MM/GBSA.

EISA-Score: Element Interactive Surface Area Score for Protein–Ligand Binding Affinity Prediction

“…Unfortunately, the problematic parameter choices of these physical models have overshadowed the valuable information extracted from the molecular surface. There are some recent efforts to directly incorporate the surface area descriptors to capture the protein–ligand potency. , However, conventional surface area models do not portray crucial physical and chemical interactions such as noncovalent bonds, hydrogen bonds, van der Waals interactions, etc., which lead to discouraging results and limited capacity to handle diverse biomolecular data sets. These issues call for robustness and scalable surface area representations for biomolecular structures.…”

“…57 However, the role of the surface area in capturing the crucial physical and chemical interactions in the biomolecular structures is not fully explored. Despite the recent efforts to integrate the surface area information into predictive models such as Cyscore 9 and GLXE 26 for protein− ligand binding affinity prediction, those surface area-based models are far from the competitive level with their counterparts.…”

Scoring Functions for Protein-Ligand Binding Affinity Prediction Using Structure-based Deep Learning: A Review

“…Thanks to the recent developments in ML force fields (Unke et al, 2021), accurate alchemical free energy calculations based on such force fields are starting to appear (Rufa et al, 2020;Wieder et al, 2021). ML-based corrections to conventional free energy calculations will also play an important role in reaching good prediction accuracy of protein-ligand binding free energies (Dong et al, 2021). While such methods are outside the scope of this review, we believe the exploration and development of ML and DL methods in the field of free energy calculations will provide very interesting outcomes in the coming years, by getting the methodology closer to chemical accuracy while significantly reducing computational costs.…”

Systematic Improvement of the Performance of Machine Learning Scoring Functions by Incorporating Features of Protein-Bound Water Molecules

“…The performances of ML-based SFs are highly dependent on the input features. Generally speaking, input features can be specific energy features, − protein–ligand atom pairwise counts or potentials, − interaction fingerprints, − mathematical features, − grid-based features, − graph-based features, ,− etc. Unlike other machine learning algorithms, many deep learning-based SFs can automatically extract features and use them for training. , Currently, extensive efforts still rely on the use of traditional ML to improve the scoring power of SFs.…”