Date of Award
5-16-2024
Publication Type
Thesis
Degree Name
M.Sc.
Department
Computer Science
Keywords
Cancer Subtyping;Contrastive Clustering;Deep Clustering;Multi-modal;Multi-omics Data
Supervisor
Ziad Kobti
Abstract
The diversity and complexity of cancer pose significant challenges in developing targeted treatment approaches, due to the disease's broad spectrum of variations across different types and even within the same type. The identification of cancer subtypes aims to detect patients with distinct molecular profiles, which influences the specific subtype's cellular makeup, disease progression, and response to treatment, thereby enabling more effective diagnosis, prognosis, and treatment of cancer. With recent advancements in technology, there's been a significant increase in the availability of multi-omics data, which is instrumental in the understanding of different cancer subtypes. However, accurately subtyping cancer through the integration and analysis of multi-omics data is challenging due to the inherent characteristics of the dataset, such as high dimensionality and significant heterogeneity. Current research typically combines multi-omics data into a single dataset through simple concatenation and then employs deep learning models to derive a lower-dimensional representation, neglecting the unique distributions of different omics data types. Additionally, they separate representation learning and sample clustering into two stages, initially learning latent representations and then applying traditional clustering algorithms, which leads to suboptimal results due to overlooking the intrinsic clustering structures in the initial learning phase. To address these limitations, we propose a novel deep unsupervised learning model, that employs multi-modal architecture with decoupled contrastive clustering to create an end-to-end clustering framework. Evaluated across eight TCGA cancer datasets, the proposed model exhibits state-of-the-art performance, surpassing existing clustering methods, with its efficacy further corroborated by survival and clinical analysis outcomes.
Recommended Citation
Herath, Achini, "Multimodal Contrastive Clustering: Deep Unsupervised Learning Approach for Cancer Subtype Discovery with Multi-omics Data" (2024). Electronic Theses and Dissertations. 9468.
https://scholar.uwindsor.ca/etd/9468