Andrew J. Bordner, Ph.D.

The general research focus of our laboratory is computational structural biology. We are developing new computational methods for modeling; predicting the structures and interactions between biomolecules, and applying these tools to discover novel reagents and therapeutics, as well as to better understand the molecular basis for diseases. More specifically, our current research interests are the following:

• Predicting membrane protein structures and interactions. Many membrane proteins are medically important as approximately 40 percent of all drugs target membrane proteins, particularly G protein coupled receptors (GPCRs). Computational methods for modeling GPCR structures are useful because only a few high-resolution experimental structures are available. Experimental evidence indicates that GPCRs often function as dimers or higher order oligomers. We are currently investigating general computational techniques for predicting membrane protein interactions and using these to design transmembrane peptides that block GPCR interactions. We are working closely with experimental colleagues in this endeavor.
• Developing structure-based prediction methods to assess which peptides bind to a particular MHC allotype. MHC molecules fall into two classes: class I MHC, which primarily bind intracellular protein fragments (peptides), and class II MHC, which bind extracellular peptides. T-cells subsequently recognize non-self peptides and initiate an immune response. These peptide epitopes may be from atypically expressed proteins in cancer or viral proteins (for class I MHC) or from bacterial proteins (for class II MHC). The binding of such peptides to MHC is required for a T-cell immune response. The knowledge of which peptides bind to particular MHC types can be used as a basis for discovering vaccines against pathogens and cancer. However, the vast number of both MHC types and protein fragments precludes comprehensive experimental determination of peptide-MHC binding affinities. Computational methods, such as the ones we are studying, can rapidly identify candidate epitopes for subsequent experimental verification.
• Applying machine learning techniques to predict the structures of protein and protein-ligand complexes. Structural information on such interactions aids in understanding the complexes's biological function and is potentially a basis for the structure-based discovery of therapeutics that modulate these interactions. Previously, we have developed such methods to predict protein-protein binding sites, to predict protein complexes through docking, to identify biologically relevant protein complexes in experimental X-ray structures, and to predict the locations of metal ion and small molecule binding sites in proteins. We plan to extend these methods to predict the structures of membrane protein complexes. Also, in collaboration with other researchers, we intend to investigate docking methods for accurately predicting the structures of protein complexes with flexible subunits.