Prediction of protein binding to DNA in the presence of water-mediated hydrogen bonds

We extend our previous analysis of binding specificity of DNA-protein compl exes to complexes containing water-mediated bridges. Inclusion of water bri dges between phosphate and base, phosphate and sugar, as well as proteins a nd DNA, improves the prediction of specificity; six data sets studied in th is paper yield correct predictions for all base pairs that have two or more hydrogen-bonds. Beside massive computation, our approach relies highly on experimental data. After deriving protein structures from DNA-protein compl exes in which coordinates were established by X-ray diffraction techniques, we analysed all possible DNA sequences to which these proteins might bind, ranking them in terms of Lennard-Jones potential for the optimal docking c onfiguration. Our prediction algorithm rests on the following assumptions: (1) specificity comes mainly from direct hydrogen bonding; (2) electrostati c forces stabilise DNA-protein complexes and contribute only weakly to spec ificity since they occur at the charged phosphate groups; (3) Van der Waals forces and electrostatic interactions between positively charged groups on the protein and phosphates on DNA can be neglected as they contribute prim arily to the free energy of stabilisation as opposed to specificity.