Zhang, Laizhi
(2024)
PROMINENT: AN INTERPRETABLE DEEP LEARNING METHOD TO PREDICT PHENOTYPES USING DNA METHYLATION.
Master's Thesis, University of Pittsburgh.
(Unpublished)
Abstract
DNA methylation (DNAm) is an epigenetic mark that regulates genome function and is important for normal development in mammals. To understand the complex interplay between genetics, DNA methylation, and phenotype, computational approaches have emerged to uncover DNAm marker-phenotype associations, often employing regression techniques, such as Methylation Risk Scores (MRS) or methylation profile scores (MPS). Recently, a deep neural network (DNN) method, MethylNet was proposed to predict traits based on DNAm array data based on variational auto encoder. However, understanding the association between DNAm data and phenotypes in the DNN architecture remains challenging since MethylNet does not directly incorporate the relationship between methylation and genes. Also, its high demand for computation time necessitates meticulous tuning of important model parameters to achieve optimal performance, making it difficult to optimize data training. To overcome these challenges, we introduce Pathway Information-based Methylation Neural Network (PROMINENT). PROMINENT incorporates gene-level DNA marker information and gene pathway priors to enhance prediction accuracy and interpretability. In a comprehensive comparison using asthma, idiopathic pulmonary fibrosis (IPF), and first-episode psychosis (FEP) data, PROMINENT shows promise in predicting phenotypes using DNA methylation data with reasonable computational time.
Share
Citation/Export: |
|
Social Networking: |
|
Details
Item Type: |
University of Pittsburgh ETD
|
Status: |
Unpublished |
Creators/Authors: |
|
ETD Committee: |
|
Date: |
16 May 2024 |
Date Type: |
Publication |
Defense Date: |
18 April 2024 |
Approval Date: |
16 May 2024 |
Submission Date: |
25 April 2024 |
Access Restriction: |
2 year -- Restrict access to University of Pittsburgh for a period of 2 years. |
Number of Pages: |
41 |
Institution: |
University of Pittsburgh |
Schools and Programs: |
School of Public Health > Human Genetics |
Degree: |
MS - Master of Science |
Thesis Type: |
Master's Thesis |
Refereed: |
Yes |
Uncontrolled Keywords: |
DNA Methylation, Deep Learning |
Date Deposited: |
16 May 2024 19:20 |
Last Modified: |
16 May 2024 19:20 |
URI: |
http://d-scholarship.pitt.edu/id/eprint/46319 |
Metrics
Monthly Views for the past 3 years
Plum Analytics
Actions (login required)
![View Item View Item](/style/images/action_view.png) |
View Item |