Speed kills, as the slogan says, and in computers what it kills could be disease.
Argonne National Laboratory researcher Andrew Binkowski’s calculations of protein structure help find ligands – smaller molecules – that attach to them, to deliver drugs that stop dangerous infections.
But without supercomputers it could take months to model a single ligand, even using the most advanced algorithms available, Binkowski says.
On Intrepid, Argonne’s IBM Blue Gene/P supercomputer, it takes just hours to evaluate the same ligand. That’s important because Binkowski and his colleagues have a database of about 5 million compounds to sift in search of the right ligand. Starting next year, Mira, IBM’s next-generation Blue Gene Q supercomputer, will be able to do such calculations as much as 20 times faster and five times more efficiently.
“We need to look at how atoms can be affected by minute changes in position and find meaningful ways to measure them” in the atom protein and ligand complex, Binkowski says. A lack of computing power meant researchers had to be somewhat imprecise, setting artificial boundaries or ignoring some interactions. Intrepid lets researchers strip away those loose restrictions and run more complete models.
Binkowski’s team has an allocation of 10 million processor hours on Intrepid, awarded through DOE’s INCITE (Innovative and Novel Computational Impact on Theory and Experiment) program. The researchers are in the midst of the largest application of highly advanced and demanding FEP/MD-GCMC (free energy perturbation distributed molecular dynamics-grand canonical Monte Carlo) computations ever performed on protein-ligand interactions. And for the first time they are applying FEP/MD-GCMC as a screening tool. That lets researchers rank different compounds’ binding free energy – a factor that reveals just how effectively a potential ligand would stick to a piece of protein.
“This is a valuable, time-saving scale by which we can measure and compare different compounds to the same target,” Binkowski says. “Whichever ones have the most favorable free energy ranking are the ones we are most interested in.”
For years, the long time it takes to run the FEP/MD-GMGC calculations limited its use, “but now we’re using it to look at thousands of ligands, which has never been done before,” Binkowski says. “It would have taken literally several lifetimes to do.”
A fourth-year fellow probes cloud droplets to boost climate predictions. Read More
A UC Davis fellow combines computing, the corn genome and growing data to help farmers… Read More
A Computational Science Graduate Fellowship recipient monitors threats to satellites in Earth’s ionosphere by modeling… Read More
A computational sciences fellow models COVID-19 virus variants and examines how people weigh complex decisions. Read More
A Colorado State fellow employs machine learning for climate modeling, putting provenance behind predictions. Read More
A LANL statistician helps cosmologists and epidemiologists grasp their data and answer vital questions. Read More