Automated molecular mechanics parameterization with simultaneous utilization of experimental and quantum mechanical data

A general set of procedures for automated parameterization of molecular mechanics force fields is presented. The current implementation for MacroModel force fields can easily be adapted to other programs running under Unix. Several variants of numerical Newton–Raphson and simplex methodologies are used to fit simultaneously both experimental and quantum mechanical data. In a sample parameterization of ethane, 24 different modes of combining various sources of reference data are compared. The best cost/performance ratio for generating quantum mechanical source data was obtained with the B3LYP/6-31G* method. The best optimization strategy consists of initial subset optimizations with a modified simplex method, followed by Newton–Raphson optimization using Lagrange multipliers. © 1998 John Wiley & Sons, Inc. J Comput Chem 19: 1146–1166, 1998