Refined models for computer calculations in protein engineering. Calibration and testing of atomic potential functions compatible with more efficient calculations.

A reappraisal has been made of interatomic potential functions for protein structure calculations using the all-atom approximation (except CH, CH2 and CH3, which are treated as "united atoms"). Some key problems are identified and treated. The potential functions are somewhat novel in form and consistent with more efficient and robust folding algorithms. In addition, the potentials are calibrated for for the rigid geometry approximation, since use of fixed standard bond lengths and valence angles (and fixed trans planar peptide groups) reduces the number of conformational variables and saves a great deal of computer time. Though these algorithms demand the use of potential functions of this special type, these functions can be readily implemented in more classical programs for the conformational analysis of proteins. They are calibrated or tested against a large body of experimental data, including extended basis set ab initio, quantum mechanical calculations, nuclear magnetic resonance spectroscopic data and dipole moment data for di- and oligopeptides, characteristic ratio data for random coil homopolypeptides, extensive data from peptide solubility studies, and experimental structures of polyalanine fibres and globular proteins. This paper will form the basis of a further report, which will include investigations of how water might be more realistically represented subject to the computing power available.