Accelerating AutoDock Vina with Containerization
TimeWednesday, July 2510:45am - 11am
DescriptionMolecular docking is an important step in the drug discovery process and uses computing resources to simulate how molecules bind together. While searching for potential drugs, often millions of molecules are tested, requiring hundreds of thousands of CPU hours. AutoDock Vina is a frequently-used, open-source molecular docking program. In addition to the source code being available, a precompiled binary is also available. The precompiled version of AutoDock Vina runs significantly slower than the program runs if it is compiled on the target hardware. However, because the current version of AutoDock Vina requires outdated compilers to compile some of the equally outdated libraries it depends upon, it is difficult and time-consuming to compile. Thus, most users of AutoDock Vina likely run the precompiled version and require a great deal of extra computational time to get their results. We attempted to fix this issue by creating a Singularity container with an environment where AutoDock Vina is easily compiled, allowing it to gain the performance benefits of being compiled without the significant effort required to compile it from source. We tested our container on the San Diego Supercomputer Center’s cluster, Comet, and found the version of AutoDock Vina in the Singularity container required 33% less (user + system) time and 33% less real time than the precompiled version of AutoDock Vina to screen a set of molecules.