To our knowledge, this is the first study to report binding free energy benchmarks on a cross-company collection of proprietary datasets. We couldn't have done it without the hard work of our amazing partner companies!

Across a total of 1,700 ligands, OpenFE shows robust performance, generates reproducible results, and achieves both sufficient throughput and rapid convergence.
Our protocol:
- ranks compounds about as well as commercial solutions
- does not require fine-tuning settings for each system

Our docs have you covered! We have a "cookbook" demonstrating how to use our API for a variety of tasks in setting up free energy calculations. docs.openfree.energy/en/latest/co...

You can find more information starting on slide 20 in our February 2025 update presentation: zenodo.org/records/1486...

To address these challenging cases, the Open Force Field Intitiative provides a tool, “BespokeFit,” for generating custom parameters for the specific chemistry of a target molecule. These custom (or “bespoke”) parameters can be used in RBFE calculations in OpenFE.

Our default force field, OpenFF “Sage,” works well on a wide variety of organic molecules relevant to drug design. However, projects involving more exotic functional groups may be limited by low force field accuracy.

This is such cool work. Ivan from the Chodera Lab is implementing a protein mutation protocol in OpenFE, and we can't wait to see what incredible studies people do with it!

It's great to see this work in print! It's not on our roadmap this year, but in the longer term we are interested in optimizing replica spacing and will look closely at your method.

But the ability of Boltz-2 to predict affinity opens up new possibilities for drug design workflows, like the Affinity Funneling approach that David Pearlman describes here.

medium.com/@dapscience/...