Galaxy Australia - making use of national and commercial GPGPU resources

Galaxy Australia is a highly successful, nationally-funded data analytics service. Life science researchers use Galaxy's best practice tool and workflow offerings, supported by the Galaxy Australia team, who deploy bioinformatics tools based on researcher demand. Galaxy tool requests from researchers have recently been dominated by tools that use general purpose graphical process units (GPGPUs). GPGPUs are particularly well-suited to some bioinformatics applications because their massive parallel computing capacity and architectural support for machine learning frameworks speed up processing, and in some cases have enabled new approaches to answering biological questions. For example, Galaxy Australia has provided the Google DeepMind 3D protein prediction tool Alphafold2 through GPGPUs on Microsoft Azure, which has been used by over 250 researchers to run more than 16,000 protein structural predictions. Other recent researcher-driven requests include tools for genome annotation, long-read sequencing alignment and mass spectrometry based protein identification. To support these requests and more proactively prepare Galaxy Australia for future demands for GPGPU-based tools, we have expanded our computational footprint to include Nectar GPGPU resources, initially through the QCIF node. In this presentation, I will discuss the change in demand for GPGPU-based tools from life science researchers, the technical challenges faced by the Galaxy Australia team to provide access to them, and the scientific gains that they have enabled.