ALCHEMD: Bridging Accessibility and Accuracy in Automated Relative Binding Free Energy Workflows

Area of Science:

• Computational chemistry
• Drug discovery
• Molecular modeling

Background:

• Alchemical free energy perturbation (FEP) is accurate for predicting drug-protein binding affinity.
• FEP adoption is limited by high computational costs and technical complexity.

Purpose of the Study:

• Introduce ALCHEMD, an automated platform for relative binding free energy (FEP-RBFE) calculations.
• Enable FEP-RBFE predictions on desktop workstations using commodity GPUs.

Main Methods:

• ALCHEMD integrates automated preprocessing, Common Structure Mapping, Combined-Structure FEP, and Convergence-Adaptive Roundtrip algorithms.
• The platform utilizes intelligent algorithms for ligand selection, parametrization, and adaptive sampling.

Main Results:

• ALCHEMD achieves 10-20 drug-protein binding predictions daily on GPUs.
• Benchmark tests show comparable accuracy (MUE = 0.86 kcal/mol) with significantly reduced simulation times (29.3 ns/pair).

Conclusions:

• ALCHEMD democratizes FEP calculations, making them accessible beyond expert groups.
• The platform accelerates drug discovery workflows through efficient and automated binding affinity predictions.