Automated and scalable ABFE workflows for Industrial-grade drug discovery R&D
Large scale deployment of molecular simulation in drug discovery R&D requires industrial-grade processes able to handle with minimal errors diverse datasets. Alchemical Absolute Binding Free Energy (ABFE) calculations are increasingly being adopted to investigate protein–ligand interactions. Although ABFE protocols were proposed over two decades ago, their industrial application requires considerable optimisations.
A team of scientists from the pharmaceutical companies Recursion and Exscientia, along with academics from the University of Vienna, recently published a preprint describing their collaborative work to deploy robust ABFE workflows using BioSimSpace over cloud computing environments such as AWS. They proposed technical improvements that reduced the frequency of simulation crashes caused by suboptimal protein–ligand restraints. Additionally, they developed an optimized lambda schedule that demonstrated lower statistical errors compared to a previously widely cited literature protocol, while also being tuned for efficient use of cloud resources.
Benchmarking across four different protein–ligand series showed that the optimized “RXRX” protocol significantly improved ABFE accuracy and precision. This protocol is now being used in production drug discovery pipelines at Recursion. OpenBioSim was pleased to provide technical support for the use of BioSimSpace in this project.

Link to the preprint
https://chemrxiv.org/engage/chemrxiv/article-details/67f78d376dde43c90839da7d
Written by Julien Michel


