U.S., British scientists develop automated program to assist in estimating physical properties from molecular simulation

An automated framework can help scientists in predicting the properties of materials from molecular simulations and is important in making it faster and easier to understand biological and chemical phenomena, according to a research team.

A group of U.S. and British scientists working in the Open Force Field Initiative consortium, an industry-funded effort, describes a new open source program to aid in optimizing force fields on the molecular level in a preprint of the journal ChemRxiv, Aug. 5.

The program is called the Open Force Field Evaluator, and it automates the rapid generation of "high- quality force fields for small molecules." The purpose of the Open FF Evaluator is to aid "insight into a range of biological and chemical phenomena and as a tool for predicting the physicochemical and mechanical properties of substances."

A vast amount of data is available in open sources, such as the National Institute of Standards and Technology ThermoML Archive, the authors point out, along with quantum chemical data. "The quantity of such experimental data in open data archives alone would require an onerous amount of human and computer resources to both curate and estimate manually, especially when estimations must be made for numerous sets of force field parameters."

They also note that the data not only are vast, but also are stored in different formats and are duplicative, corrupted and sometimes not machine readable. All of these elements mean that it takes a huge amount of time to manually process the data for use in simulations.

The new automated Open FF Evaluator is able to automatically estimate properties from simulation data and can make use of an application program interface to estimate new properties. The researchers compare the results of the Open FF Evaluator in two situations, one measuring "a test set of binary density and enthalpy of mixing measurements" and another against "a training set of density and enthalpy of vaporization measurements."

A scalable and efficient framework

The researchers describe the two core philosophies of the framework's design.

"The first is that the framework should be readily scalable for any required calculations from running on a single machine up to running across hundreds of high performance compute nodes, and potentially even into the cloud," the authors said. "Secondly, it is constructed so that every aspect is user extendable via a flexible plugin system." 

The Open FF Evaluator automatically can select and extract appropriate data for a particular task from different open data sources and then "benchmark the broad spectrum of commonly used force fields." It can also make use of new properties and be integrated into other software. Users can customize the way they use the program.

Among the physical properties that the framework covers are mass density, dielectric constant, enthalpy of vaporization, enthalpy of mixing, excess molar volume and solvation free energy. Plug-ins can add new physical properties if necessary.

The authors stress that the Open FF Evaluator eliminates redundancies in data, and decreases the number of simulations to be performed and analyzed. They also link to where the framework can be obtained and how to get help for problems in using it.

Simon Boothroyd et al., "The Open Force Field Evaluator: An automated, efficient, and scalable framework for the estimation of physical properties from molecular simulation," ChemRxiv, 05 August 2021. https://chemrxiv.org/engage/chemrxiv/article-details/610ad0ed45805d722f80e4de