OpenMM secures federal funding though an NIH NIGMS R01 grant

/
nigms-logo.jpeg

Recently, OpenMM applied for NIH funding to seek a sustainable federal source of support to continue to serve and adapt to the changing needs of the molecular simulation community by providing a fast, flexible, and extensible platform for advanced biomolecular simulations.

We’re thrilled to report that the NIH has awarded us funding through Mar 2025 via NIH grant R01GM140090. Funding will continue to support lead OpenMM developer Peter Eastman, as well as new developers based in the computational biophysics lab headed by Gianni de Fabritiis. Together, this will enable us to not only continue to support, optimize, and maintain OpenMM, but to also extend it to take advantage of the unfolding revolution in quantum machine learning potentials that continue to transform our field.

With a newly redesigned website, a newly-established OpenMM Consortium helping steer scientific directions, recruitment of science communicator Joshua Mitchell to lead a major effort to refine our documentation and materials, added support for GPU-accelerated pytorch and tensorflow based potentials, and migration to the conda-forge ecosystem (with 158K downloads from conda-forge this year already), we’re off to a great start!

To read more about our plans to continue to extend OpenMM to tightly integrate OpenMM with modern ML frameworks such as TensorFlow, PyTorch, and JAX; allow machine learning potentials or collective variables defined in these machine learning frameworks to be easily used within OpenMM; and and Python libraries to make it easy to build next-generation hybrid quantum machine learning / molecular mechanics (QML/MM) models within these frameworks, check out our NIH research proposal here.

Thanks to all of you who submitted letters of support! Your support means the world to us.

Securing sustainable funding to enable OpenMM to continue to power the next decade in biomolecular modeling and simulation

/

OpenMM is the most widely-used open source GPU-accelerated framework for biomolecular modeling and simulation. It has been cited more than 1300 times, downloaded over 280,000 times from conda-forge alone, and has run on more than one million distinct computers. Its Python API makes it widely popular as both an application (for modelers) and a library (for developers), while its C/C++/Fortran bindings enable major legacy simulation packages to use OpenMM to provide high performance on modern hardware. OpenMM has been used for probing biological questions that leverage the $16B global investment in structural data from the PDB at multiple scales, from detailed studies of single disease proteins to superfamily-wide modeling studies and large-scale drug development efforts in industry and academia.

Originally developed with NIH funding by the Pande lab at Stanford, we now aim to fully transition toward a community governance and sustainable development model and extend its capabilities to ensure OpenMM can power the next decade of biomolecular research, guided by the OpenMM Consortium. To fully exploit the revolution in QM-level accuracy with quantum machine-learning (QML) potentials, we also plan to add plug-in support for QML models augmented by GPU-accelerated kernels, enabling transformative science with QM-level accuracy. To enable high-productivity development of new ML models with training dataset sizes approaching 100 million molecules, we will develop a Python framework to enable OpenMM to be easily used within modern ML frameworks such as TensorFlow and PyTorch. Together with continued optimizations to exploit inexpensive GPUs, these advances will power a transformation within biomolecular modeling and simulation, much as deep learning has transformed computer vision.

Recently, we applied for federal funding to realize this vision via a new NIH Focused Technology Research & Development R01 proposal, with strong support from the biomolecular simulation community. You can read the scientific components of the proposal we submitted here: [PDF]

We welcome your feedback on how OpenMM can continue to serve the needs of the biomolecular simulation community over the next decade and beyond!

OpenMM 7.1 beta released

/

Christmas comes early! We've released a beta of OpenMM 7.1, packed with speed improvements and new features, including:

  • Optimized clang builds of both Anaconda packages and ZIP installers, offering anywhere from a 30% to 50x boost for some applications that use the CPU platform.
  • Custom Forces can now compute energy derivatives with respect to global parameters! Lambda dynamics can now be implemented via a CustomIntegrator.
  • Gay-Berne ellipsoid potential!
  • Bonded forces can now use periodic boundary conditions.

To get the updated OpenMM conda package, use the beta channel:

conda install -c omnia/label/beta openmm==7.1.0

If you have already been using the dev channel 7.1.0 nightly builds, force a downgrade first:

# Force downgrade to 7.0.1
conda install --yes -c omnia openmm==7.0.1
# Clear local cache
conda clean -plti --yes
# Install the beta
conda install --yes -c omnia/label/beta openmm==7.1.0

See the OpenMM SimTK page for more information on the beta release. Feel free to give feedback there, or on the OpenMM GitHub issue tracker.

Nightly dev builds are now called 7.2.0. You can always get the latest version with:

conda install --yes -c omnia/label/dev openmm

CECAM Workshop on developing interoperable and portable molecular simulation software libraries

/

I'm excited to be participating in the CECAM Workshop on Developing Interoperable and Portable Molecular Simulation  Software Libraries this week at the Forschungszentrum Jülich, organized by Julien Michel, along with Christopher Woods, Peter Eastman, and Gareth Tribello.

Slides and materials used during the workshop can be found here:

  • Examples and materials: https://github.com/choderalab/cecam-2015-julich-workshop