UNIVERSITY PARK, Pa. — When Wen Jiang, Dorothy Foehr Huck and J. Lloyd Huck Chair for Structural Biology and professor of biochemistry and molecular biology, saw a need to give his graduate students a unique hands-on experience using Cryogenic Electron Microscopy (Cryo-EM) imaging techniques for his BMMB 531 Biomolecule Structure class, he reached out to the technical team at the Institute for Computational and Data Sciences (ICDS) to integrate two complex software items into Penn State’s Roar supercomputer.
The software, Cryo-EM Single Particle Ab-Initio Reconstruction and Classification (CyroSPARC) and REgularised LIkelihood OptimisatioN (RELION), allows users to take 2D images of proteins, viruses and other biological assemblies, and convert them into 3D structures.
“The Cryo-EM technique is used in the structured biology field, and there are typically large amounts of image data that come from the electron microscope,” Jiang said. “Very often we have terabytes of data that we need to convert and it’s an extensive task. Both CryoSPARC and RELION are powerful tools that can be used in this field.”
Jiang said that in more recent years, researchers have relied on using CryoSPARC, not only for its fast performance, but also because of the use of a front-end web browser interface and back-end use that makes use of artificial intelligence tools and graphics processing units (GPUs), which help process data and help construct the models.
Incorporating CryoSPARC into Roar was no easy task, according to Adam Focht, ICDS systems engineer.
“There were a number of ways we could go about integrating CryoSPARC and RELION into Roar, and a number of challenges we faced because we have limitations and the software allows users to access both a front- and back-end that interact with one another,” Focht said.
The ICDS technical and special projects teams worked together with other higher education institutions to understand how the software works, making sure security requirements are met and ensuring researchers have minimal barrier to entry.
To do so, the technical team separated the front- and back-end interfaces in the class’s cluster to not only allow for more effective and efficient use of the web browser application, which is used for preparing research and processing, and the back-end where the science, compute and analysis is happening. This “cross-block” set up allowed the front- and back-end to be set up as specialized zones for different parts of the computing process.
“This whole process was unique and I was very excited about putting it together,” Focht said.
For his class, Jiang wanted to provide students with a hands-on experience in image analysis and 3D reconstruction, in addition to his regular lectures – an experience Jiang says is not always readily available to students.
The students were tasked with selecting a publicly available dataset that included peer-reviewed articles in the biomolecule structure field, download it into the class cluster on Roar, import it into the CryoSPARC software and complete a full image analysis of molecules. The goal of the class was to reproduce the same quality reconstruction of a 3D model as was reported in the peer-reviewed article from dataset of the student’s choosing.
“I think this experience was very valuable to the students because many of them would probably never have had the opportunity to use this software on a research project,” Jiang said. “In class lectures, we teach students about science, theory and principles, but I believe in the full learning experience that includes both principal level understanding and building practical, hands-on skills. Having the CyroSPARC software installed on Roar filled that gap.”
Focht added that having CryoSPARC and RELION software available on Roar not only enhanced the classroom experience for these students but also has the potential for opening it up to the broader research community at Penn State.
“This class kickstarted the interest in implementing complex software onto Roar but also gave us the motivation to get working on how it could also be beneficial to researchers across the University,” Focht said.
Jiang added that the class was used almost as a pilot study, allowing both Jiang and the ICDS technical team to learn what does and doesn’t work, what users are experiencing when using the software on Roar and what areas could be improved for broader use.
“This setup on Roar is beneficial both for education and research needs,” Jiang said. “We want to expand this beyond classroom needs and make this resource available for researchers, at any experience level, to gain experience in advanced image processing software on computing clusters and to use this tool for cutting-edge research projects.”