NSF CC* PA Science DMZ Research Collaboration Speaker Series
Date: Thursday, October 30
Time: 12:00 p.m.–12:45 p.m.
Location: Zoom
On Thursday, Oct. 30, from 12 to 12:45 p.m., Daniela Fera, associate professor and department chair of chemistry and biochemistry at Swarthmore College, will present “Connecting Microscopes to Computation: How Shared Infrastructure Accelerates Structure Determination” as part of the National Science Foundation (NSF) CC* PA Science DMZ Research Collaboration Speaker Series.
The event will take place over Zoom.
Abstract: We seek to determine 3D models of complexes relevant to the immune response. Specifically, we use cryo-electron microscopy (cryo-EM) to visualize how antibodies recognize viruses and how kinases regulate cellular signaling. To resolve these structures at atomic-level resolution, we collect large datasets, often 5–10 terabytes, from a single session on a transmission electron microscope at the University of Pennsylvania. Using the PA Science DMZ cyberinfrastructure, we move raw image files directly from Penn to Swarthmore’s new Firebird high-performance computing cluster, developed jointly with Lafayette College. Efficient data transfer and analysis are critical. This high-speed, secure connection would enable near real-time feedback during data collection and tightly integrate data acquisition with downstream analysis. Once on Firebird, GPU-intensive workflows using cryoSPARC and machine learning tools such as Topaz are used to construct 3D maps of our biological complexes. Topaz uses convolutional neural networks trained to recognize molecular particles directly from micrographs, and is efficient for large and heterogeneous samples. However, its performance can be limited for smaller or less distinct complexes—such as the kinase assemblies we are now targeting—where the signal-to-noise ratio is low and particle features are subtle. These challenges motivate ongoing development of training datasets and network architectures to extend cryo-EM analysis to more difficult targets. Overall, the PA Science DMZ regional cyberinfrastructure has connected researchers, cryo-EM instruments, and high-performance computing to enable rapid transfer and analysis of large datasets, and will expand access to additional structural biology studies to accelerate scientific discovery.