Authors:

Lucia Zhang

Date Created:

Not specified

Course Title:

Professor:

Not specified

About Paper:

Cryo-electron microscopy (cryo-EM) is widely used to characterize biological molecules, ranging from proteins to drug delivery vehicles. However, analysis of these particles remains predominantly manual, creating costly bottlenecks and limiting scalability. Furthermore, while correlative light and electron microscopy (CLEM) provides subcellular localization information by combining fluorescence and electron signals, existing workflows for image correlation are similarly labor-intensive. In this project, we aimed to build a modular workflow to address these challenges. The first component is an automated, high-throughput pipeline for the detection and characterization of liposomes, which serve as a model system for drug delivery mechanisms. I trained a U-Net convolutional neural network to segment liposomes from electron micrographs. I then developed MATLAB and Python scripts to extract quantitative morphological features, such as size and circularity, from the segmented liposomes. The second component of the workflow improves the speed and precision of CLEM image registration. I implemented a novel two-step correlation strategy by first aligning large-scale structural features (such as cryo-grid hole patterns) and then refining the alignment using smaller fluorescent fiducials. The liposome characterization pipeline rapidly recognizes particles and extracts accurate, population- level statistics. Maximizing the number of features used for CLEM alignment additionally enables more robust correlations while increasing efficiency. Together, these tools offer enhanced resolution in CLEM imaging workflows, along with high-throughput, spatially contextualized analysis of liposome populations. This framework provides a foundation for applying CLEM to therapeutic delivery mechanisms for preclinical development. Keywords: Cryo-EM; Machine Learning † Presenting Undergrad Author; ‡ Contributing Undergrad Author; * Undergrad Acknowledgment

Source:

Purdue University / 2025

Topics:

No topics listed

Co-authors:

Lucia Zhang