Graduate Studies

 

First Advisor

Jitender Deogun

Second Advisor

Yanbin Yin

Degree Name

Doctor of Philosophy (Ph.D.)

Department

Computer Science

Date of this Version

12-5-2023

Document Type

Dissertation

Comments

Copyright 2023, Minal Khatri

Abstract

This study investigates two critical areas in bioinformatics: enhancing transparency in medical image analysis and advancing the discovery of Anti-CRISPR (Acr) proteins, which have potential in developing more precise and controlled CRISPR-Cas gene editing tools. While CNN’s are increasingly applied in critical fields like medical diagnosis, understanding their decision-making process remains a challenge. Although visualization techniques like Saliency maps offer insights into CNN’s decision-making for individual images, they do not explicitly establish a relationship between the high-level features learned by CNN’s and the class labels across dataset. To bridge this gap, Formal Concept Analysis (FCA) framework is leveraged as a image classification model, establishing a novel method for understanding the relationship between abstract features and class labels in medical imaging. The model’s performance is validated across a range from the simpler MNIST dataset to more complex histopathological image datasets like Warwick-QU and BreakHIS. Simultaneously, the study explores the genomic context of Acr genes and the 3D structure of Acr proteins for Acr discovery, which are not extensively explored in current bioinformatics tools. By leveraging genomic context, we overcome data scarcity and develop a machine learning model capable of discovering new Anti-CRISPRs. Additionally, the 3D structure analysis aids in developing machine learning classifiers to classify proteins by Acr type and CRISPR-Cas systems. Overall, this research makes significant contributions to the field of bioinformatics, by developing robust methodologies, enhancing our understanding of medical image analysis and advances Acr protein discovery.

Download

Share

COinS