Research Mentor(s): Jaie Woodard, Postdoctoral Fellow
Research Mentor School/College/Department: Computational Medicine and Bioinformatics, Michigan Medicine
Presentation Date: Thursday, April 22, 2021
Session: Session 5 (3pm-3:50pm)
Breakout Room: Room 11
The goal of my research was to prepare my own dataset, which includes a training set and a test set, and build my own neural network in order to predict protein backbone with cryo-EM density map as my input. In order to get to know more about deep learning and neural networks, I completed 2 coursera courses recommended by my mentor to gain the skills needed to start building my own neural network by using python software. I also read one major literature that focused on cryo-EM technologies to further learn about how incorporating density maps of proteins can provide us with important structural information of proteins. A density map can be obtained using cryo-electron microscopy in order to obtain atomic-resolution models of the protein, including the coordinates of the backbone protein atoms. However, the index/connection of these atoms are lost during the process of obtaining the density map. Predicting the protein backbone will allow us to restore the connectivity of the backbone atoms in the protein and improve the CR-I-TASSER, which incorporates both deep learning and I-TASSER force fields in order to provide accurate structures of the protein backbone.
Authors: Jeonghoon Hyun
Research Method: Computer Programming