The mission of Discovering Dengue Drugs - Together - Phase 2 is to identify promising drug candidates to combat the Dengue, Hepatitis C, West Nile, Yellow Fever, and other related viruses.

The extensive computing power of World Community Grid will be used to complete the structure-based drug discovery calculations required to identify these drug candidates.

On August 12, 2011, Dr. Robert Malmstrom gave this lecture on the Discovering Dengue Drugs Together project running on World Community Grid.

November 14, 2012

Hello! Some users have expressed interest in the details of our Phase 2 (free energy of binding; FEB ) calculations. I am happy to provide this information and briefly discuss some of my recent work with parameter optimization, code stability, and accuracy testing.

Our initial FEB calculations, using an early version of the CHARMM code, produced an unacceptable number of incomplete (i.e failed) calculations when tested on a large scale with World Community Grid. This problem was traced to instabilities in the molecular dynamics simulations associated with fixed boundary conditions being selected for variable-sized ligands. To correct this problem, we ported the latest version of the CHARMM code from charmm-gui to World Community Grid, modified the automated boundary parameter selection routine to accommodate large ligands, and improved the accuracy of the stereochemical parametrization of bound ligands.

To test the robustness and accuracy of these FEB calculation code enhancements, we selected co-crystal structures for eight very different proteins in the Protein Data Bank. Each ligand-protein complex had highly accurate binding data from microcalorimetry experiments. Our new code and associated optimized parameter set produced stable molecular dynamics simulations for the eight tested ligand-protein systems. Moreover, component energy terms extracted from molecular dynamic simulations converged during the timescale of the simulation and were highly reproducible for multiple replicate calculations. Finally, calculated FEB values were in good agreement with experimental measurements (e.g., for the AmpC protein, the calculated FEB was -8.0±2.5 kcal/mol while the experimental binding energy was measured as -8.2 kcal/mol).

Now that we have a robust and stable code, with parameters that generate molecular dynamics simulations with reproducible energy convergence, we will re-examine Phase 1 results to better discriminate between false positive and true positive hits.


More. . .