Background
Current docking and design software often fail to model the flexibility seen in proteins or ligands with many rotatable bonds. Also, while many complexes involve multiple ligands, cofactors, and ions, current software can consider only one ligand at a time. Understanding the ruling principles whereby protein receptors recognize, interact, and associate with molecular substrates and inhibitors is of paramount importance in drug discovery efforts. As a consequence, a number of protein-ligand docking programs have been developed that provide a means to predict how a putative drug will react with its target protein, or how a mutation in a protein might affect its interaction with a small molecule. Despite major advances over the last three decades, current docking programs still face challenges, one of which is the lack of accurate potential functions which evaluate and rank potential ligands from chemical collections. In particular, the current RosettaLigand scoring function still suffers from limitations, especially an inadequate description of electrostatic interactions. In RosettaLigand, electrostatics are crudely represented by two terms; a pair potential using a distance-dependent Coulomb model and an orientation-dependent hydrogen bonding potential, which was not specifically parameterized for protein/small molecule interfaces, but instead generated from protein-protein complexes. Additionally, RosettaLigand ignores the polar character of π-electrons and hydrogen atoms in aromatic rings such that important short-range electrostatic interactions (e.g., cation – π, π – π and lone-pair (lp) – π interactions) are not represented. While a full Poisson-Boltzmann calculation would lead to a more accurate view of the electrostatic environment, implementing such a procedure remains computationally intractable. Thus, a model is needed which can provide a more accurate description of short-range electrostatic interactions within protein/small molecule interface while still maintaining efficiency.
Inspired by the idea that short-range electrostatic interactions have partial covalent character, i.e., caused by sharing electrons, we plan to derive a novel short-range electrostatic potential based on geometrical parameters of pairs of orbitals observed in a non-redundant protein-small molecule database. This scheme features to produce more accurately defined geometries for short-range electrostatic interactions in terms of orbital overlapping. In addition, it retains pairwise decomposability which is critical for speed. The new electrostatic potential will be benchmarked through the recovery of short-range electrostatic interactions using different datasets. We will then test it for docking decoy recognition and binding affinity prediction. The performance of the newly developed orbital-based short-range electrostatic potential will be compared to those of Rosetta’s existing hydrogen-bonding potential and the “action center”-based pair potential.
Alumni Project Members: Steven Combs, Thomas Willcock