Calculation of the Lowest Energy Configurations of Nucleotide Base Pairs on the Basis of an Electrostatic Model

Abstract

In this report we have taken as a point of departure the idea that the principal force acting between two nucleotide bases is the Coulomb law force between the localized formal charges on each atom of one base and each atom of the other base. The charges were computed by semiempirical quantum‐mechanical methods for both σ and π electrons separately and added together. The energies of attraction between all possible pairs of the four common bases adenine, uracil, cytosine, and guanine in all possible coplanar orientations were computed automatically on a Honeywell 800 digital computer. Twenty‐seven deep potential‐energy minima were found. Although it was not possible to find gas‐phase geometries and energies with which to compare these computed stable states, the energies in solution and known geometries in crystals, cocrystals, and polymer fibers were, with a single possible exception, completely consistent with the predicted structures and relative energies. Examination of the computed structures showed that, with few exceptions, the N–H···O and N–H···N short contacts are essentially linear, so that although the forces which led to these geometries were purely electrostatic in nature, the complexes may for convenience be thought of as being ``hydrogenbonded complexes.'' In several of the computed structures, nonlinear N–H···O and N–H···N close contacts were observed and these can be conveniently thought of as ``bifurcated hydrogenbond complexes'' which have been observed in nature although not, as yet, in purine and pyrimidine complexes. The need for employing formal charges to reproduce the electrostatic forces was demonstrated by showing that dipole—dipole forces do not lead to potential minima and/or stable structures. Tests with a variety of computed charge densities showed that variation in the charges does not affect the equilibrium geometries or their relative stabilities although it does affect the absolute energies. The method described therefore appears to be a satisfactory method for obtaining trial geometries for x‐ray‐diffraction analyses of crystals and fibers, and for estimating relative stabilities for use in predicting which pairings can occur where choice is possible, i.e., in codon—anticodon complexes, in hairpin‐folded RNA complexes, as well as in cocrystals.