(169dl) The Molecular Simulation Design Framework (MoSDeF): Enabling High-Throughput Simulations via Active Learning Integration Workflows

The Molecular Simulation Design Framework (MoSDeF)¹ was developed as a suite of tools to mitigate challenges associated with initializing and conducting molecular simulations, with a focus on facilitating reproducibility.² By providing standardized, open-source libraries, MoSDeF can automate steps in the preparation of chemical/biological systems for simulation, thereby minimizing unnecessary human involvement and associated errors. Furthermore, the automation of workflows enables easier setup of high-throughput screening processes, through integration with workflow management tools such as signac and signac-flow.³ To further leverage and extend these types of workflows, a new package, genGrouper, has been developed for generalized chemical representation of simulation inputs, utilizing the PyTorch deep learning library.⁴ Additionally, progress has been made in supporting features necessary for more complex water models, such as TIP4P/2005f and TIP5P models, and a standard screening workflow centered around MoSDeF software has been performed and made available for facile extension to more sophisticated high-throughput active learning approaches.

References

1. “MoSDeF” [Online]. Available: https://mosdef.org
2. Cummings, P. T. et al. (2021). Open‐source molecular modeling software in chemical engineering focusing on the Molecular Simulation Design Framework. In AIChE Journal (Vol. 67, Issue 3). Wiley.
3. Adorf, C. S., et al. Simple data and workflow management with the signac framework. Comput. Mater. Sci., 146(C):220–229, 2018.
4. Ansel, J., et al., (2024) PyTorch 2: Faster Machine Learning Through Dynamic Python Bytecode Transformation and Graph Compilation. In ASPLOS '24: Proceedings of the 29th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, Volume 2. ACM.