Prediction of the histidine-95 pKa perturbation in triosephosphate isomerase using an electrostatically trained neural network (SONNIC)

Abstract

A backpropagation neural-network trained with AM1 SCF-MO derived molecular electrostatic potentials (MEPs) for a set of 30 substituted imidazoles of known ionisation constant can be used to predict the pK_a of imidazole under the influence of the α-helix dipole found for histidine-95 in triosephosphate isomerase; the results agree quantitatively with experimental evidence that this residue has an anomalous value such that its catalytic nature is unconventional.