Parmscan is an automatic engine for force-field parameterization. In this work, we applied both systematic search (SS) and a genetic algorithm (GA) to optimize the force-field parameters (bond length, bond angle, as well as torsional angle terms) to reproduce the relative energies of conformational pairs as well as other molecular properties such as vibrational frequencies. We present an example of how to apply Parmscan to reproduce the relative energies of 11 hydrocarbons by optimizing the torsional parameter of Csp3-Csp3-Csp3-Csp3 using both systematic search and the genetic algorithm. Both of the two methods successfully found the lowest RMS error, which is 0.51 kcal/mol (the unsigned mean error is 0.37 kcal/mol). The Cornell et al. model (Parm94, ref. 21) achieves an RMS error of 0.78 kcal/mol. A second example is the application of the genetic algorithm to optimize the torsional parameter Csp3-O-Csp3-O and bond angle parameter O-Csp3-O simultaneously for 11 dioxanes, to reproduce the experimental relative energies. After 300-400 iterations of GA optimizations, the RMS deviation is reduced to 0.56-0.57 kcal/mol, slightly better than that of Parm99 (ref.22) and much better than that of the Cornell et al. model. In further applications, the bond length and bond angle parameters of hydrocarbons and benzene were optimized to reproduce the experimental vibrational frequencies. Encouraging results were obtained compared to the Cornell et al. force field: for the low vibrational frequencies of ethane, propane, and butane, the new model achieves an unsigned mean error of 19 cm-1, compared to 30 cm-1 of the Cornell et al. model; for 20 vibrational frequencies of benzene, the new model can also give much smaller unsigned mean error (29 vs. 53 cm-1). It seems that after appropriate parameterizations, a simple harmonic molecular model, such as employed in AMBER, can also reproduce lower vibrational frequencies of molecules quite well.
- Force-field parameterization
- Genetic algorithms
- Systematic search
ASJC Scopus subject areas
- Safety, Risk, Reliability and Quality