Consistent force fields for carbohydrates were hitherto developed by extensive optimization of potential energy function parameters on experimental data and on ab initio results. A wide range of experimental data is used: internal structures obtained from gas phase electron diffraction and from x-ray and neutron diffraction, vibrational frequencies, dipole moments, unit cell dimensions and lattice energies. The range of model compounds covered so far includes alkanes, ethers, alcohols, ketones and mono-and disaccharides. Electrostatic interactions are handled by fractional charges assigned to individual atoms. Charges are modeled such that Mulliken population analyses are reproduced. Morse functions are used for all bonded interactions; experimentally derived dissociation energies are used as parameters. Van der Waals interactions are modeled with Lennard-Jones 12-6 functions. The anomeric and exo-anomeric effects are accounted for without addition of specific terms. The work is done in the framework of the Consistent Force Field which originated in Israel and was further developed in Denmark. The actual methods and strategies employed have been described previously. Extensive testing of the force field is reported, and ways and means of improvement are indicated. Principles of mapping of conformational space are discussed, and a discussion on which properties to preferentially reproduce in modeling is invited.
SHORT HISTORY OF THE CFF ATTEMPTSWe have worked on the Consistent Force Field 3 for thirty years, 4 and tried to apply the method to saccharides for twenty; the previous history has been published. 5 The full power of the CFF was not available in its updated form until 1985, 6 and then optimization 789 T 790 RASMUSSEN was undertaken. In 1991 we finished the first serious bid for an optimized force field for carbohydrates and their derivatives, PEF91L. 7Until then we had worked with hand-fitted force fields like most other researchers in modeling, and our main results were the demonstration that relaxation in all internal degrees of freedom is necessary for meaningful calculation of conformations of disaccharides and, by implication, of oligosaccharides; and the proof that quite good results can be provided even with rather primitive potential energy functions, if their parameters are judiciously chosen.PEF91L was applied to a variety of problems, notably the charting of the conformational space of gentiobiose. 7Our work over the years had shown that it was of overriding importance to model the non-bonded interactions adequately. We developed a new and more rational strategy of optimization, 8 involving heavy optimization on crystal structures, which resulted in our best potential energy function so far, with the parameter set PEF95SAC. This has been documented in much detail. 9The present paper records further tests of the force fields, and directions for improvement are indicated.