The binding sites of proteins include smaller regions called hot spots that are major contributors to the binding free energy, and hence are crucial to binding any ligand at that site. Hot spots can be determined by screening libraries of fragment-sized organic molecules for binding to the target protein by X-ray crystallography or nuclear magnetic resonance (NMR). The computational mapping algorithm FTMap, developed at Boston University and a related program, Atlas, by Acpharis are faster analogs of experimental screening techniques. Application of computational mapping to fragment based ligand discovery (FBLD), which is a goal of this proposal, requires comparison of the predicted pose of a ligand fragment and poses of mapping probes in important hot spots of a protein binding site. The procedure involves three preliminary steps: (1) identification of binding hot spots, (2) generation of the accurate poses of molecular probes in the main hot spots, (3) prediction of the binding mode of the ligand fragment. Steps 2 and 3 depend on the accuracy of the energy function used in the fast Fourier transform (FFT) based algorithm of computational mapping. Here we propose development of structure based statistical potentials (SBSPs) focused on the main hot spots as a straightforward way to improve specificity of mapping energy function. SBSPs are constructed as energy values derived from the frequencies of atomic contacts in the databases of known protein-ligand structures. They rely on statistics of interaction and reference states of the system of interacting atoms with assigned atom types. Our approach will use non-specifically binding docking decoys as a reference state (DARS). We will derive the interaction states from the hot spot regions of the binding site. For construction of SBSPs we will select training, reference, and validation sets of protein-ligand structures from the PDBbind database, which collects complexes with known experimental binding affinities. We will rely on the Atlas software of Acpharis for parameterization of the fragments and for docking to generate the decoys. We will define atom types to reflect distinct chemical properties of an atom in its environment. To construct the potentials we will calculate maximum interaction distances for each atom type pair from the radial distribution function of interactions. To test, the new potentials correlation between the statistical pairwise energy and experimental binding energy will be calculated for the small molecule validation set. We will add a pairwise DARS term, based on statistical potentials into the mapping energy function and reweight energy terms to fit the experimental binding energy. We will evaluate the docking performance of the new energy function on molecules from the validation set. Proposed developments will increase the accuracy of virtual screening by computational mapping for the selection of fragment hits, reducing or even removing the need for experimental screening in fragment based ligand discovery.
Public Health Relevance Statement: The goal of this proposal is to use structure based statistical potentials to increase the accuracy of virtual screening by computational mapping for the selection of fragment hits, reducing or even removing the need for experimental screening in fragment based ligand discovery.
Project Terms: Accounting; Affinity; Algorithms; analog; Atlases; base; Base Pairing; Binding; Binding Proteins; Binding Sites; Boston; chemical property; Chemicals; cofactor; Complex; Computational algorithm; Computer software; Data Set; Databases; density; Derivation procedure; Development; Docking; Environment; Fourier Transform; Free Energy; Frequencies; Generations; Goals; Hot Spot; improved; Libraries; Ligands; Linear Regressions; Molecular Conformation; Molecular Probes; Molecular Weight; Nuclear Magnetic Resonance; Performance; Piper; Procedures; programs; Protein Fragment; Proteins; Radial; Sampling; screening; Series; Set protein; Shapes; Site; small molecule; Specificity; statistics; Structure; System; Techniques; Testing; Training; Universities; Validation; virtual; Weight; X-Ray Crystallography
