Fast automated placement of polar hydrogen atoms in protein-ligand complexes

Hydrogen bonds play a major role in the stabilization of protein-ligand complexes. The ability of a functional group to form them depends on the position of its hydrogen atoms. An accurate knowledge of the positions of hydrogen atoms in proteins is therefore important to correctly identify hydrogen bonds and their properties. The high mobility of hydrogen atoms introduces several degrees of freedom: Tautomeric states, where a hydrogen atom alters its binding partner, torsional changes where the position of the hydrogen atom is rotated around the last heavy-atom bond in a residue, and protonation states, where the number of hydrogen atoms at a functional group may change. Also, side-chain flips in glutamine and asparagine and histidine residues, which are common crystallographic ambiguities must be identified before structure-based calculations can be conducted.

We have implemented a method to determine the most probable hydrogen atom positions in a given protein-ligand complex. Optimality of hydrogen bond geometries is determined by an empirical scoring function which is used in molecular docking. This allows to evaluate protein-ligand interactions with an established model. Also, our method allows to resolve common crystallographic ambiguities such as as flipped amide groups and histidine residues. To ensure high speed, we make use of a dynamic programming approach.

Our results were checked against selected high-resolution structures from an external dataset, for which the positions of the hydrogen atoms have been validated manually. The quality of our results is comparable to that of other programs, with the advantage of being fast enough to be applied on-the-fly for interactive usage or during score evaluation.