Computational methods have advanced our understanding of bio-molecular binding processes and have become a vital tool in structure-based drug design. Computational docking techniques in particular, utilize free energy functions and algorithms, to rapidly produce a quantitative ranking of likely binding sites of small molecules to larger target molecules based on binding free energy. AutoDock is a widely used molecular docking program that has been shown to be reasonably accurate predicting the binding sites of ligands (small molecules) with 6 or fewer rotatable bonds. However, accuracy decreases and becomes problematic after 10. We are interested in finding ways to expedite this docking procedure. Our studies, which have focused on protein flexibility, shows promise as the foundation for a new screening protocol that would permit the accurate determination of binding sites of larger ligands, while utilizing cost effective computational methods such as the  improved accuracy and speed of Autodock Vina, a reliable new alternative to AutoDock.

A snapshot from our research comparing the experimentally known structure (red) of

a cancer drug bound to the onco-protein MDM2 to AutoDock’s predicted binding site (magneta)

In addition, we employ accelerated Molecular dynamics simulation programs such as NAMD (Nanoscale Molecular Dynamics) that use classical Newtonian physics to study the time dependant structure, dynamics, and thermodynamics of biological molecules. The microscopic properties of atomic positions and velocities can be translated into macroscopic quantities such as temperature, pressure and volume using statistical mechanics. This allows us to determine the flexible binding site residues of the target protein most associated with these binding interactions. These side chain protein residues can then be made flexible during docking runs in order to better simulate changes in physical structure associated with ligand binding.

The acceleration is made possible by the use of Graphic Processing Unit (GPU)s that speed up parts of the MD computational tasks:

NVIDIA's GPU WORK STATION at MSU

Recenty, we have developed a simple approach to predict relevant rotatable bonds for a molecular docking.  This approach allows us to quickly identify the pertinent residues from the protein that would greatly affect the effectiveness of the docking of screened drug candidates. See these examples in the Materials Genome section.