QuantumBind-RBFE: Accurate Relative Binding Free Energy Calculations Using Neural Network Potentials

2025-01-03

AI Summary

Insufficient accuracy in predicting drug–protein binding affinities hinders computational drug discovery. To address this, we propose a novel relative binding free energy (RBFE) method based on the AceForce 1.0 neural network potential (NNP), the first application of this TensorNet-architecture NNP to RBFE calculations. Our approach supports all chemical elements and charged drug molecules, and enables stable molecular dynamics simulations with a 2-fs time step—over twofold faster than comparable NNPs. On standard benchmark datasets, our method significantly outperforms GAFF2 and ANI2-x, achieving accuracy on par with OPLS4, while maintaining high computational efficiency. The implementation—including source code and pre-trained models—is publicly released. This work establishes a scalable, physics-informed, AI-driven computational paradigm for high-fidelity drug design.

Technology Category

• Machine Learning: Matrix & Tensor Methods
• Humans and AI: Other Foundations of Human Computation & AI
• Knowledge Representation and Reasoning: Computational Complexity of Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Social Networks and Social Media: Computational social science

Abstract

Accurate prediction of protein-ligand binding affinities is crucial in drug discovery, particularly during hit-to-lead and lead optimization phases, however, limitations in ligand force fields continue to impact prediction accuracy. In this work, we validate relative binding free energy (RBFE) accuracy using neural network potentials (NNPs) for the ligands. We utilize a novel NNP model, AceForce 1.0, based on the TensorNet architecture for small molecules that broadens the applicability to diverse drug-like compounds, including all important chemical elements and supporting charged molecules. Using established benchmarks, we show overall improved accuracy and correlation in binding affinity predictions compared with GAFF2 for molecular mechanics and ANI2-x for NNPs. Slightly less accuracy but comparable correlations with OPLS4. We also show that we can run the NNP simulations at 2 fs timestep, at least two times larger than previous NNP models, providing significant speed gains. The results show promise for further evolutions of free energy calculations using NNPs while demonstrating its practical use already with the current generation. The code and NNP model are publicly available for research use.

Problem

• Drug Discovery
• Protein-Ligand Binding
• Computational Methods

Innovation

• QuantumBind-RBFE
• Neural Network Potential
• Drug Discovery Acceleration