(63e) The Open Force Field Initiative: Better Force Fields through Open, Data-Driven Science

Physical modeling of biologically relevant molecular entities is a powerful tool for predicting biomolecular properties and function, interpreting biophysical experiments, and designing new small molecules or biomolecules with therapeutic utility. While molecular simulation has benefited from recent massive increases in computational power, improvements in molecular mechanics force fields have lagged behind. As a result, the utility of physical modeling is limited by the uncertain accuracy and restricted domain of applicability of current force fields. Despite a revolution in data availability and ubiquity of computing resources, we lack technologies to draw on the wealth of available experimental and quantum chemical data to rapidly and systematically improve force fields.

The Open Force Field Initiative (OpenFF) is collaboration focused on the development of 1) extensible, open source toolkits for constructing, applying, and evaluating force fields; 2) the curation of public datasets necessary to build high-accuracy biomolecular force fields; and 3) the generation of improved molecular force fields for biomolecular and other soft matter applications. In this talk, I will discuss recent efforts by OpenFF to simplify the existing force field framework, put force field development in an explicitly data-driven context, and gather the data needed to make this effort possible, as well as the current status of our initial force field parameterization effort. An important part of our goal is to make force field development reproducible and automated, and I give an update on progress towards that end. I will also describe the creation of the Open Force Field Consortium, a joint industry-academic initiative to help advance OpenFF aims, as well as the longer term plans of OpenFF, how we hope it will be useful to you, and how you can join the effort.