Force fields are used in a variety of research fields including computer-aided drug design, biomaterials, and polymer chemistry. However, force fields also continue to limit the accuracy of predictions of physical properties. Current parameterization of these force fields involves a huge amount of human effort -- often years of work -- and depends heavily on the chemical intuition of those involved. The Open Force Field Initiative is working to replace this tedious process with an automated machinery to learn parameters and chemical perception. Our new SMIRKS-based force field format, SMIRNOFF, allows all parameter types to be defined independently. This allows for easier extension compared to the traditional atom type-based force fields where the chemical perception of all parameter types is intertwined.
We will need to be capable of programmatically learning SMIRKS patterns in order to fully automate force field parameterization. In this work, we present ChemPer -- a new tool for generating SMIRKS patterns based on clustered fragments (i.e. bonds, angles, or torsions) which should be assigned the same force field parameter. We demonstrate the utility of ChemPer by clustering fragments based on a reference force field, and then regenerating those parameters starting with a simple set of alkanes, ethers, and alcohols. Next, we create SMIRKS patterns for a protein SMIRNOFF which match the parameters from AMBER99. We conclude with a discussion of other potential applications and expansions to ChemPer.