In the Carlson group, our studies broadly focus on the molecular recognition between ligands and proteins, from the fundamental biophysics of ligand binding to applied inhibitor design. Through studying protein-ligand interactions, our research can help the scientific community to better understand biomolecular processes and develop effective therapeutics. Because we use computational methods to probe biological processes, our research is inherently multidisciplinary. We frequently collaborate with researchers both on and off campus. This provides my students with a broad exposure to Medicinal Chemistry and Biophysics beyond our own computational techniques. Our research is funded by the NIH, the NSF, and the Beckman Foundation.
Developing New Tools for Structure-Based Drug Design: We are refining a technique for combining multiple protein structures (MPS) in structure-based drug discovery. A few methods have been proposed that combine MPS. Ours was the second method presented in the literature, and, to date, it is the only one to be experimentally verified.
Our initial goal is to use MPS to incorporate protein flexibility into drug design. We are using well-characterized protein systems to refine the method into a robust tool for the biomedical community. We have initiated our studies with molecular dynamics (MD) simulations of unbound (apo) HIV-1 protease. Our method uses snapshots from those MD simulations as MPS for the design of receptor-based pharmacophore models. The resulting models were highly selective for known inhibitors of HIV protease and effectively weeded out non-inhibitors in database searches. We are excited that the use of MPS allowed us to identify the key features of bound ligands from an unbound protein structure. This is particularly dramatic because HIV-1 protease undergoes a large rearrangement upon binding inhibitors.
Developing a Comprehensive Protein-Ligand Database (Binding MOAD -- Mother of All Databases): We have created the largest database of protein-ligand complexes. Other databases of protein-ligand complexes have been limited to a few hundred entries. These are compiled in a ground-up fashion, adding new entries as they are gleaned from the literature. We are using a top-down approach to gather all relevant entries from the Protein Databank (PDB). After the 2005 update, Binding MOAD has 8219 entries. BLAST searches were used to divide the dataset into related protein structures, resulting in 2723 unique protein systems. After painstakingly searching the crystallography literature, we have collected binding affinity data for 2193 complexes.
We are also developing software to examine underlying principles of the basic biophysics behind protein-ligand recognition. For instance, we have created a program that analyzes any bound ligand to determine the volume of the cavity, volume of the ligand, and buried surface area. While this might sound trivial, it is not. Our program works for any protein-ligand complex whether it is a closed cavity or open cleft, whether there are bridging water molecules, whether there are multiple ligands in the site, etc. We have started to use this software with our database to determine general characteristics of tight binding such as the degree of buried surface and the volume/volume ratio of the ligand to binding site. This information should be useful in drug design for creating an optimal complement to the binding site.
Our database work is synergistic with our MPS project. In processing the database with BLAST, we inherently produce a supplemental database of the homologous entries (redundancies) in the PDB. We plan to use this subset of data with our MPS method. It may be possible to use an entire set of homologous proteins, binding various ligands, to create models that identify broad-spectrum inhibitors. This approach may be advantageous in the development of antibiotics and antivirals where activity against many pathogens is desirable.