FlowerMD: Flexible Library of Organic Workflows and Extensible Recipes for Molecular Dynamics

Abstract

flowerMD is a package for reproducibly performing multi-stage HOOMD-blue (Anderson et al., 2020) simulation workflows. It enables the programmatic specification of tasks including definition of molecular structures, forcefield definition and application and chaining together simulation stages (e.g., shrinking, equilibration, simulating a sequence of ensembles, tensile testing, etc.) through an extensible set of Python classes. The modular design supports a library of workflows for organic macromolecular and polymer simulations. Tutorials are provided to demonstrate package features and flexibility. Statement of need High-level programmatic specifications of molecular simulation workflows are needed for two reasons. First, they provide the information necessary for a simulation study to be reproduced, and second, they help lower the cognitive load associated with learning and performing simulations in general. Reproducible simulations benefit the research community by enabling studies to be validated and extended. Lowering the cognitive load of performing molecular simulations helps computational researchers of all levels of expertise reason about the logic of a simulation study. This is particularly important for researchers new to the discipline because developing the tools needed to perform experiments often involves: (a) gaining new software development skills and knowledge, and (b) repeating work that others have already performed. Recent advances in open-source tools have made the programmatic specification of molecular simulation components easier than ever (Anderson et al. A. P. Thompson et al., 2022; M. Thompson et al., 2023). Individually, each of these tools lower the cognitive load of one aspect of an overall workflow such as representing molecules, building initial structures, parameterizing and applying a forcefield, and running simulations. However, stitching these pieces together to create a complete workflow presents a need that we address in the present work. The computational researcher who follows best practices for accurate, accessible and reproducible results may create a programmatic layer over these individual software packages (i.e. wrapper) that serves to consolidate and automate a complete workflow. However, these efforts often use a bespoke approach where the entire workflow design is tailored toward the specific question or project. Design choices might include the materials studied, the model used (e.g. atomistic or coarse-grained), the source of the forcefield in the model, and the simulation protocols followed. As a result, this wrapper is likely unusable for the next project where one of the aforementioned choices changes, and the process of designing a workflow Albooyeh et al. (2023). FlowerMD: Flexible Library of Organic Workflows and Extensible Recipes for Molecular Dynamics. Journal of Open Source Software, 8(92), 5989. https://doi.org/10.21105/joss.05989.