DEMM: A meta-algorithm to predict the pka of ionizable amino acids in proteins

Abstract

The protonation states of ionizable amino acid residues often have a direct influence on the functioning of a protein. The acid dissociation constant (in logarithmic scale, pKa) of these residues is hence an important determinant of protein function. To predict pKa, we integrated two complementary state of the art pKa prediction methods, DEPTH and microenvironment modulated screened Coulomb potential approximation (MM-SCP). The performance of the integrated predictor, DEMM, was benchmarked on a dataset of 47 residues with experimentally measured pKa values. DEMM has an average prediction error of < ~0.5 pH units and was statistically significantly superior to the DEPTH and MMSCP methods. The method’s utility is enhanced by its speed, accuracy and its applicability to proteins of varying sizes.