Non-linear Multi Omics Data Integration Method Using Conditional Variational Autoencoders

Advances in technology have enabled the study of diseases through multi-omics data, which combines information from genome, epigenome, transcriptome, proteome, and metabolome levels. Unlike single-omics approaches that provide limited insights, multi-omics integration offers a comprehensive understanding of biological systems by capturing interactions across molecular layers. In recent years, several methods have been developed to integrate omics data. For example, Simidjievski et al., 2019 introduced techniques that use Variational Autoencoders (VAEs) for data integration. Similarly, Zarayeneh et al., 2017 proposed a method called the Integrative Gene Regulatory Network (iGRN), which combines multiple layers of omics data using a network made up entirely of gene nodes. This thesis focuses on developing data integration architectures based on conditional variational autoencoders (CVAEs). The key advantage of this approach is that it allows class label information to be incorporated during the data integration process. To the best of our knowledge, CVAEs have not been applied in previous multi-omics research. Additionally, new methods for integrating more than two datasets using CVAEs have been introduced. This is a novel contribution to the field of multiomics data integration, as no prior studies have explored the use of CVAEs for integrating multiple datasets in this context. The proposed architectures were tested on both real and simulated datasets. The results from both studies showed that adding an outcome variable (class labels) to regular VAEs improved predictive performance. Additionally, integrating data from multiple datasets produced better results compared to using a single dataset for predictions or using VAEs without incorporating labels.

Name thesis:degree_name

Master of Science (MSc)

Discipline thesis:degree_discipline

Mathematics & Statistics

Grantor dc:publisher.institution

Graduate Studies

Year dc:date.issued

2025

Author dc:creator

• Gustinna Wadu, Dimuth Adeepa Gunarathne

Advisors dc:contributor.advisor

• Chekouo, Thierry
• Wu, Jingjing

Committee members dc:contributor.committeemember

• Kopciuk, Karen
• Aminghafari, Mina
• Forkert, Nils

• Conditional Variational Autoencoder
• Multi-Omics Data

Statement dc:rights

• University of Calgary graduate students retain copyright ownership and moral rights for their thesis. You may use this material in any way that is permitted by the Copyright Act or through licensing that has been assigned to the document. For uses that are not allowable under copyright legislation or licensing, you are required to seek permission.

Language dc:language.iso

en

DOI

https://dx.doi.org/10.11575/PRISM/48306

OAI identifier oai:identifier

oai:ucalgary.scholaris.ca:1880/120697