- Ultra large scale docking for ligand discovery.
- Development and testing of a receptor desolvation method in docking. The scientific question is the role of accounting for water in our ligand discovery campaigns. There has been much focus on waters in docking over the past 3 decades, which has intensified with the development of Inhomogeneous Solvation Theory (IST). To test the theory in docking for ligand discovery at atomic resolution, we incorporated a grid-based IST scoring term into DOCK v3.7. We perform retrospective docking to 26 systems and prospective testing on the model cavity in Cytochrome C Peroxidase gateless mutant. Two key findings emerge: i. the water displacement energy is on average small, contributing little to the DOCK score. Thus, the impact over many molecules is slight, as is seen through the small differences in the majority of poses and ranks when comparing screens with and without grid-based IST and in the slight improvements of retrospective enrichment results (over the 26 systems). ii. there is a meaningful impact on a small number of molecules: GIST is able to prioritize or deprioritize certain molecules (none of the three deprioritized molecules bought were shown to bind). In the prospective study, we, compellingly, identified a molecule that was predicted to make a water mediated interaction, which was confirmed in subsequent x-ray structures. The ligand's imidazole nitrogen interacts with a water which interacts with Asp 233. We are currently examining faster methods for calculating the displacement of water during docking and applying these methods to other biological systems including beta-lactamase and the HER family, on which I will focus in my future research.
Figure. Docked pose (magenta) compared to subsequently determined crystallographic pose (green). Water-mediated interaction shown.
- Exploring protein conformational response to congeneric ligands. The scientific question is the role of protein conformational changes in the recognition of a series of ligands. Medicinal chemists will often make a series of small modifications to ligands, but little is known about how the protein might change to accommodate these ligands. When we searched through the PDB we found few cases of congeneric complexes. To answer this question, we turned to an engineered site in T4 lysozyme, a congeneric series of eight ligands, and x-ray crystallography. Matt Merski solved 8 structures of T4 lysozyme using multistate crystallography of a homologous series of ligands. We clustered these structures of T4 lysozyme and observed three discrete states of the protein in response to the eight congeneric ligands (benzene to hexylbenzene), rather than a gradual conformational response to the ligands.
Figure. Drawing by Trent Balius in colored pencil. Depicts the the incamental growth of the small molecule from benzene to hexylbenzene and the conformational change of the alpha-helix in the closed state to a 3-10 helix in the open state. The stairs represent the discrete nature of the protein response.
- As a graduate student with Robert Rizzo at Stony Brook University, the overarching scientific objective I pursued was applying and developing computational tools to impact drug discovery. 1. We used molecular dynamics simulations in conjunction with free energy calculations and footprint analysis to elucidate the molecular- basis of resistance to EGFR drugs. We compared the L858R mutant to the L858R&T790M resistant mutant (Figure). We observe there is a reduction in the bridging water between erlotinib and the protein residues T790 and T854. A key finding is that reduction of the water network in part explains the resistance of T790M to erlotinib and other ligands. This illustrates the importance of solvation effects in receptor-ligand binding. 2. We implemented a new scoring function into DOCKv6.5 based on per-residue decomposition of the interaction energy between ligand and receptor, termed molecular footprint. The footprint signatures of two molecules may be compared (using distance or correlation) and used to describe the similarity of the two molecules. We rescore docked molecules by comparing their signatures with a reference molecule. This allowed us to identify molecules that "feel" the same to the receptor although they are chemically different. We have used this method to rescore results from several virtual screens to good effect. I contributed to development of tools that have impacted ligand discovery and I will continue to do so in my future research.
Figure. Water in two sites mediate interactions between erlotinib and mutant EGFR proteins L858R (left) and L858R&T790M (right).
Figure. Representative molecular footprints for two different ligands derived from per-residue decomposition of the intermolecular van der Waals interactions as a function of primary sequence. Similarity is quantified using Pearson coefficient (r) and Euclidean distance (d).
