Using BlueGene to characterize protein ligand interactions with DOCK and
Structure-based design can significantly reduce the time and expenditure
required for drug discovery. However, origins of ligand binding which drive
molecular recognition are difficult to predict and remain a challenging
problem. We will present strategies, examples, and results using the tools
of all-atom docking and molecular dynamics simulations to characterize
binding on the BG/L platform. DOCK program setups and options used for
compiling the MPI version of the program on BG/L and BG/P will be discussed
including tests for docking a database of 100,000 small organic molecules to
test MPI scaling for virtual screening. Protocols, setups, and benchmarks
for using the molecular dynamics program NAMD for simulating HIVgp41 and
EGFR will also be presented.
Massively-Parallel Molecular Dynamics Simulations on New York Blue.
Molecular simulation provides a powerful tool to probe biochemical
function, and tera-to-peta Flop computational resources allow for larger
systems and timescales to be explored. In this talk, we will discuss
molecular simulation on BG/L with the NAMD software package. A detailed
description of how large-scale explicit solvent simulations have been
set up and run will be given. The key bottlenecks in parallel
performance of molecular simulations will be discussed, and scalability
of differently sized systems on BG/L will be presented. Examples will
be given from initial results on the dynamics and energetics of key
structural states of the heterotrimeric G-proteins.