From Bench to GPU: A Biologist's Journey into Virtualised Cryo-EM Data Processing on Bunya HPC

Title: From Bench to GPU: A Biologist’s Journey into Virtualised Cryo-EM Data Processing on Bunya HPC Abstract: As structural biology embraces an era of big data and GPU-intensive workflows, traditional data processing methods are no longer keeping pace, at the University of Queensland (UQ) we have managed to solve this issue at scale for our researchers utilising Open OnDemand and clever integration to Bunya HPC. In this presentation, I share a biologist’s perspective on navigating this shift from broken traditional data processing methods, to leading the field as the new gold-standard thanks to an interdisciplinary collaboration with the Research Computing Centre and the structural biology community at UQ. Together, we developed a visual, on-demand virtual desktop platform for the Bunya supercomputer that enables complete Cryo-EM data processing, no command-line skills required. This new environment integrates key software tools used in structural biology, including CryoSPARC, Relion, ModelAngelo, ChimeraX, Coot, and more, within a single virtual desktop interface. Users can process large datasets, visualize structures, and leverage A100, H100 and L40 GPUs with minimal technical overhead. By removing barriers to HPC access, this platform empowers researchers to focus on discovery, not infrastructure. I reflect on what made this project successful —clear communication across disciplines, shared vision, and a user-centred design. As compute becomes increasingly specialised, while scientific research demands higher compute, I argue that platforms like this are essential for bridging domain expertise and digital infrastructure, and I offer lessons for institutions seeking to support transdisciplinary science at scale.