Modeling chemical bonding effects for protein electron crystallography: the transferable fragmental electrostatic potential (TFESP) method

This paper addresses the problem of determining the electrostatic potential of large proteins by the superposition of potentials calculated for small fragments. The use of different atomic and molecular fragments is considered for reproducing the molecular electrostatic potential of different conformations of N-acetylalanine methylamide (NAAMA) with an acceptable degree of error as measured by conventional R factors used in crystallographic structure refinement. Three different divisions of NAAMA are tested, producing fragments that incorporate increasingly more complete descriptions of molecular bonding with diminishing accuracy in geometric fit to the parent molecule: single atoms in molecules, bonded atoms in molecules and selected functional groups, such as the backbone peptide moiety, or the alpha-carbon, beta-carbon and their associated H atoms. In the resolution range 2.5-25 Angstrom, the fairly straightforward use of single atoms in molecules reduces the calculated R factors by 5-15% over a free-atom superposition. No significant further improvement was found at the lowest resolutions with a superposition of single bonds in molecules and R factors were found to degrade with larger fragments at higher resolutions because of poor geometry fits to the atoms of the parent molecule. Because the potential distribution even for single atoms depends on the environment, the best accuracy will be obtained by using a library of fragment potentials calculated for each type of atom as a function of important protein conformations.