Authors:

Maria Munoz Perez

Date Created:

Not specified

Course Title:

Professor:

Not specified

About Paper:

Cellular proteins frequently form helical filaments, playing fundamental roles in both regulated biological processes and pathological conditions, such as Alzheimer's disease and other neurodegenerative disorders. Extensive research with cryo-electron microscopy (cryo-EM) and other techniques has greatly advanced our understanding of these helical structures. However, identifying the structure of experimental images remains challenging. A tool that can recognize whether a protein's structure has been previously identified and provide relevant data from the database would be invaluable. This research aims to develop and train a neural network for efficient image recognition of helical structures from the Electron Microscopy Data Bank (EMDB). Preprocessing the data is needed, as numerous incorrect and missing entries were identified on the database. To address this, we used helical indexing with the cylindrical projection of a 3D map (HI3D) and implemented a correlation-based validation method to clean essential helical parameters. Additionally, an automatic method for validating new database entries will be developed. The neural network, currently trained on a subset of the cleaned dataset, demonstrated effective feature extraction. Scaling up the entire helical database and data augmentation is expected to enhance generalizability for real cryo-EM images. A classifier will be designed on this pre-trained network to accurately identify structures from the database. Preliminary results suggest that this project is a promising tool for the recognition and validation of helical EMDB structures, with potential applications in improved database management and accurate protein structure identification. Keywords: Helical Structures; EMDB; Image Recognition; Neural Network

Source:

Purdue University / 2024

Topics:

No topics listed

Co-authors:

Maria Munoz Perez