FEP Protocol Builder

Free energy perturbation (FEP) calculations are a powerful tool for predicting the binding affinity of small molecules to proteins and other drug targets. However, developing accurate FEP protocols can be challenging and time-consuming, especially for complex systems.

Traditional FEP protocol development requires manual exploration of a large parameter space. Additionally, it can be difficult to determine which parameters have the greatest impact on FEP performance, resulting in overfitting.

FEP Protocol Builder (FEP-PB) is an automated workflow that uses active learning to rapidly generate accurate FEP protocols. FEP-PB iteratively selects the most informative protocol parameter settings to test, based on the results of previous calculations. This approach allows FEP-PB to quickly identify a set of parameter settings that produce accurate FEP results, with minimal human intervention.

To validate FEP-PB, we applied it to a variety of pharmaceutically relevant systems, including the previously challenging MCL1 system. FEP-PB was able to rapidly generate accurate FEP protocols for all the systems tested, with minimal human intervention.

In a real-world drug discovery setting, FEP-PB was used to generate an accurate FEP protocol for the p97 system. FEP-PB was able to generate a more accurate protocol than the expert user, rapidly validating p97 as amenable to free energy calculations.

FEP-PB is a powerful new tool that can help researchers to rapidly develop accurate FEP protocols for a wide range of systems. This could lead to the discovery of new drugs more quickly and efficiently.

Check out the full work here