-
Notifications
You must be signed in to change notification settings - Fork 56
OpenFE Roadmap
The long-term goal is to make Open Free Energy the platform of choice for flexible, performant alchemical simulation at scale — able to rank tens of thousands of compounds across the system types drug discovery actually encounters, with little hand-holding — and, in doing so, to become a hub that draws the community's contributions back into a shared ecosystem rather than scattering them. Near-term progress toward that goal includes specific projects aligned to five distinct objectives:
We are constantly working on benchmarking and validating our tools to enable us to maintain and refine the accuracy of our tools. We are also working with partner projects such as Open Force Field, to improve the quality of our models and by consequence free energy estimates. Doing this allows us to measure how OpenFE fares across different systems, and identify where adjustments to our protocols (e.g. introducing new sampling methods) will be necessary.
We continually look for opportunities to lower the GPU-hours per transformation — through collaborating to develop advanced features in MD engines, to more efficient lambda scheduling, and the elimination of redundant sampling — so that a fixed compute budget covers more ligands.
We will make it easier for users to apply OpenFE tools to a wider range of scientific endpoints by broadening applicability through new methods (e.g. SepTop and ABFEs) as well as enabling more complex systems (e.g. membrane embedded proteins).
We are focusing on building modular tools and frameworks to accelerate progress. For example, through our upcoming work on protocol continuation, we will be focusing on refactoring our protocols to make them easier to modify with minimal code adjustments. This will allow members of the community to use the OpenFE API to quickly implement new methods, for example swapping out the simulation sampler in RBFE calculations to enable different types of simulations such as non-equilibrium switching or expanded ensembles.
Our goal is to enable routine free energy campaigns that scale to tens of thousands of molecules. To achieve this, we are focusing on continually improving our user experience, adding new tools (e.g. our upcoming execution engine Exorcist, and adding the ability to resume simulations) to simplify workflows as well as making sure that users can leverage our tools as efficiently as possible (e.g. better documentation).