Feature Implementation and AI Integration for Cryo-EM Analysis through Studying Beta-Galactosidase

Ruiqi Li

Mentor: Dr. Jana Ognjenovic, Frederick National Laboratory for Cancer Research.

Date/Time: August 22nd, 2024 at 2:40pm.

Abstract: Cryo-EM, cryo-electron microscopy, is a technology that shoots the high-speed electrons as source beam through the frozen samples of cells, viruses, and other biological matters to obtain their structural information and gain deeper insights at a molecular level of observation. Two mainstream software that handle the image processing of Cryo-EM are RELION and cryoSPRAC.

Beta-Galactosidase (known as beta-gal) is an enzyme that is responsible for lactose hydrolysis to galactose and glucose. Two separate workflows were built on Cryo-EM and RELION to process the retrieved micrograph images of beta-gal. Both workflows were able to enhance the resolution of the images to ~2 Angstroms after tuning the settings. The generated maps fit the volume well in ChimeraX.

RELION is an open-sourced platform that allows researchers to implement their own programs and scripts for analysis purposes. I managed to develop an interactive program that allows a user to manually pick particles in the micrograph and output their sizes. The particle can be precisely selected by generating a circle after user confirming the two ends of it. The output sizes can be saved for future use in the user’s research.

SAM2 is a state-of-the-art model that can identify and mask objects in the picture. It has the potential to help locate and mask the particles in every micrograph. I have implemented the model, and it can take micrograph images; however, the model needs more enhancements because it missed some particles, failed to identify the projections of the particles, and mistakenly masked empty areas. This new AI method will make a great contribution to image processing and structural learning of biological matters in Cryo-EM analysis.