Our FightAIDS@Home – Phase 2 team is getting acclimated to running the research software on this very powerful yet unconventional computational resource. We are analyzing preliminary results from the first batches of research tasks, which were used to test different simulation approaches, and honing in on the most efficient ways to use the vast amount of donated computing time. We have also been preparing a large set of new work units while retooling our data processing and analysis pipelines to keep up with the massive inflow of results.
Using BEDAM simulations for FightAIDS@Home – Phase 2 presents an enormous opportunity to refine and enrich the results from Phase I. However, before refining both hits identified in Phase 1 and new compounds studied by our HIVE Center collaborators, we need to develop and test new ways to run our software under the constraints imposed by the grid computing infrastructure.
The current set of research tasks (‘work units’) was designed to serve three purposes. First, we evaluate how our new simulation schema tailored for free energy calculations using the distributed and heterogeneous resources of World Community Grid compares with the results of more traditional homogeneous high performance clusters. We have launched two different sets of simulations on World Community Grid that differ in how energy function parameters vary across many copies of each target ligand-protein complex being simulated. Second, the simulations allow us to revisit the calculation of binding free energies (binding scores) and prediction of binding modes (poses) for a well-studied dataset of HIV Integrase protein that we studied previously in the SAMPL4 binding free energy prediction challenge. By performing these different simulations on this benchmark set of complexes, we can design and optimize future batches of simulations to be as efficient as possible in terms of volunteers’ resources.
The first set of simulation results (experiments 0-52) are almost complete and have been analyzed. We anticipate that this first simulation scheme is vastly superior to the second scheme being tested in experiments 53-105. We are approximately 60% through these first 106 experiments and if preliminary analysis of the second set of experiments proves our hypothesis correct, we may choose to end the second set of simulations before completion in order to move on to new batches, all using the better simulation scheme.
Furthermore, while you’ve been crunching our work units without replication (quorum of 1), we designed the batches to have built in redundancy — which means that research tasks are assigned more than once in order to compare the results and confirm their accuracy — which we now conclude is too conservative. It is evident from analysis of the first set of batches that the current simulation protocols can be scaled down for future batches, allowing us to simulate more complexes at once on World Community Grid moving forward. We are currently preparing the next set of batches using docking results of more complexes from the SAMPL4 challenge which we recently received from the MGL laboratory at Scripps.
We anticipate that this first year of FightAIDS@Home Phase 2 will be extremely rewarding for our group and collaborators at the HIVE Center, and we’re very grateful for having been given this opportunity by IBM, and most importantly, World Community Grid volunteers.
It has also been an exciting month for the Levy lab. It’s our pleasure to announce that the January issue of Protein Science is a special issue in recognition of Ronald Levy’s contributions to the field of computational biophysics! The issue is free to read online. Professor Ron Levy, who designed the BEDAM tool used in this phase of the project, leads the molecular simulations in Phase 2 and we are very happy to help celebrate his accomplishments in this field.