MetaxFEP: Automated Absolute Binding Free Energy Calculations Workflow on MetaX GPUs

Abstract:
The widespread adoption of absolute binding free energy perturbation (ABFEP) calculations has been limited by two critical challenges: lack of end-to-end automation and excessive time-to-solution (TTS). We present MetaxFEP, an automated Python-based workflow that addresses these limitations by integrating the complete ABFEP pipeline for MetaX GPU architectures. The tool’s modular design enables fully automated execution from system preparation through free energy analysis, significantly reducing manual intervention. Validation against two established benchmarks (BRD4 bromodomain and TYK2 kinase systems) demonstrates robust prediction accuracy. When deployed on an eight-GPU MetaX server, the optimized implementation achieves remarkable computational efficiency, completing single ligandprotein pair evaluations in under 10 hours, representing a 17-fold acceleration over 128-core CPU implementations. An intelligent convergence detection feature provides additional TTS reductions with minimal accuracy trade-offs. MetaxFEP represents a significant advancement in practical ABFEP implementation, combining automation, speed, and reliability for drug discovery applications. The tool will be openly available through MetaX’s developer community (https://developer.metax-tech.com/).