Skip navigation
Please use this identifier to cite or link to this item:
Title: A Computational Approach to the Study of Regulatory Genomics
Authors: Cofer, Evan Mitchell
Advisors: Troyanskaya, Olga G
Contributors: Quantitative Computational Biology Department
Keywords: computational biology
deep learning
machine learning
regulatory genomics
Subjects: Bioinformatics
Computer science
Issue Date: 2023
Publisher: Princeton, NJ : Princeton University
Abstract: The study of transcriptional regulation is critical to furthering our understanding of life and disease. In the last few decades, high-throughput sequencing methods such as RNA-seq, ChIP-seq, and Hi-C have provided us with the means to assay different aspects of transcriptional regulation with relative ease. The availability of such data has given rise to the field of computational regulatory genomics, which seeks to completely understand all phenomena and mechanisms governing transcriptional regulation through the rigorous analysis and integration of such data. In that vein, we conducted an in-depth investigation into the 3D structure of chromatin during Drosophila embryogenesis. We demonstrate that chromatin structures are highly dynamic, and chromatin structuring elements such as boundaries fall into distinct categories associated with unique regulatory factors and distinct dynamic patterns. Because of the scale of the data being analyzed, the computational challenges in regulatory genomics are non-trivial. Therefore, a major goal of this work has been to develop novel computational techniques for modeling various aspects of transcriptional regulation. In specific, we developed DeepArk, a set of deep learning models for model species that is capable of accurately predicting regulatory features (e.g. chromatin accessibility, histone marks, transcription factor binding) directly from genomic sequences. Beyond simply developing cutting-edge methods, we also carefully demonstrate their relevance to both computational and experimental research endeavors (e.g. genome editing) in the study of regulatory genomics. We also developed Selene, a tool to ease the development and application of models like DeepArk, and created AMBIENT, an algorithm for efficiently identifying optimal neural architectures for regulatory feature prediction tasks. In summary, we have uncovered new aspects of chromatin regulation while simultaneously expanding the computational resources available to regulatory genomics researchers.
Type of Material: Academic dissertations (Ph.D.)
Language: en
Appears in Collections:Quantitative Computational Biology

Files in This Item:
File Description SizeFormat 
Cofer_princeton_0181D_14494.pdf13.16 MBAdobe PDFView/Download

Items in Dataspace are protected by copyright, with all rights reserved, unless otherwise indicated.