(595e) Application of Concepts from Modeling Integrated Computing for Molecular Simulation for Workflow Encapsulation

Authors: 
Cummings, P. T., Vanderbilt University
Klein, C., Vanderbilt University
Iacovella, C. R., Vanderbilt University
Sallai, J., Vanderbilt University
Lédeczi, Á., Vanderbilt University
McCabe, C., Vanderbilt University
Over the last few decades, the molecular simulation community has worked together to
develop a host of open-source simulation packages, as well as many transferable
force fields, substantially reducing the barrier for performing simulations. However,
despite these efforts, much of a simulation workflow (i.e., the steps required to run and
analyze simulations) are still performed using ad hoc codes, scripts and procedures,
limiting the reproducibility of simulation results by other researchers, and making it
challenging to generally perform property screening and optimization. To translate
these complex workflows into a set of tasks that can be automated, combined with other
tasks (such analysis routines) and embedded within hierarchies of application-hardened
engineering methodologies (such as stochastic optimization), we borrow concepts from
the computer science field of model integrated computing (MIC). MIC addresses the
problems of designing, creating, and evolving information systems by providing rich,
domain-specific modeling environments and tools for model analysis and synthesis.
Here, we report our efforts to develop tools to facilitate system initialization, atom-typing,
and workflow design that integrate with the ever-growing scientific Python stack.
These tools and procedures help to capture the details required to retain knowledge of
the tasks and processes used to perform a simulation; this augments reproducibility,
aids in training new researchers to perform complex simulation operations, and
ultimately enables general screening operations to be performed.