Design of Multivalent Antibodies

Artificial antibodies can be designed with multiple variable domains and expressed in mammalian cells - these can be used to specifically target toxins to cells, to simultaneously bind multiple subunits of a single protein complex, or to simultaneously bind multiple proteins (i.e. neutralize multiple toxins from an infectious disease agent. This work has been partially published - see Craig, RB et. al. (2012) in Publications. This work is in collaboration with Dr. Seth Pincus’ Lab at Children’s Hospital of New Orleans.

Modeling of the HIV gp160 Complex

This project is related to the project above : the optimal design of the linker region and orientation between anti-gp41 and anti-gp120 requires knowledge of how the gp120 epitope is oriented relative to gp41 in 3 dimensions. This work is in collaboration with Dr. Seth Pincus’ Lab at Children’s Hospital of New Orleans.

Modeling a Possible Secondary Binding Site for UbcH7-Ub onto the HECT H3 domain

Detailed enzymatic analyis of the transfer and elongation steps of the Ubiquitination pathway suggest a secondary binding site for initial binding of UbcH7-Ub the the HECT H3 domain. Using molecular modeling and docking, we propose a binding site that differs from the canonical binding site found in the published crystal structure. A paper describing this work is currently in preparation. This work is in collaboration with Dr. Arthur Haas’ Lab at LSU Health Sciences Center.

Design, Simulation, and Mechanism of Antiviral Peptides

This work is in collaboration with Autoimmune Technologies in New Orleans.

Design of Optimized Constrained Graphs for Free Energy Calculation of Drug-Lead Compounds

Alchemical Free Energy Calculations are highly computationally intensive. Therefore, planning a barrage of calculations for a large set of drug lead compounds, when carefully optimized up-front, can shave many CPU-years worth of computation (and likely weeks to months of clock time) from a calculation protocol over a large set of compounds. We have developed a highly efficient graph theoretic algorithm for minimizing the number of free energy calculations that must be performed in order to simulate free energies of binding for large compuund sets. A paper on this work is currenly in preparation. This work is in collaboration with Dr. David Mobley’s Lab at UC Davis and Schrodinger.

Parallel and Serial Tree-Based Monte Carlo Algorithms

Monte Carlo algorithms are inherently parallelizable, but their performance on optimization problems can be significantly impacted by large energy barriers. We have developed and are testing an MPI based Monte Carlo implementation in C++ that will have the ability to “peek-ahead” multiple steps (and possibly over large energy barriers) to arrive at more optimal solutions within a given search space.