Molecular Subtyping of Cancer Based on Distinguishing Co-Expression Modules and Machine Learning

Abstract

Molecular subtyping of cancer is recognized as a critical and challenging step towards individualized therapy. Most existing computational methods solve this problem via multi-classification of gene-expressions of cancer samples. Although these methods, especially deep learning, perform well in data classification, they usually require large amounts of data for model training and have limitations in interpretability. Besides, as cancer is a complex systemic disease, the phenotypic difference between cancer samples can hardly be fully understood by only analyzing single molecules, and differential expression-based molecular subtyping methods are reportedly not conserved. To address the above issues, we present here a new framework for molecular subtyping of cancer through identifying a robust specific co-expression module for each subtype of cancer, generating network features for each sample by perturbing correlation levels of specific edges, and then training a deep neural network for multi-class classification. When applied to breast cancer (BRCA) and stomach adenocarcinoma (STAD) molecular subtyping, it has superior classification performance over existing methods. In addition to improving classification performance, we consider the specific co-expressed modules selected for subtyping to be biologically meaningful, which potentially offers new insight for diagnostic biomarker design, mechanistic studies of cancer, and individualized treatment plan selection.

Phenomics and Robust Multiomics Data for Cardiovascular Disease Subtyping

“…[19][20][21][22] The nodes in these interaction networks can represent genes, proteins, or even individuals, and their connections can describe a functional relation (eg, protein binding) as well as a similarity relation (eg, patient similarity networks). Owing to its versatility, network modeling can aid at multiple stages of the computational disease subtyping workflow, ranging from the identification of subtype-specific coexpression modules, 23 to the definition of individualized molecular networks, 24 and to the integration of data across various domains to capture disease-related variability. 25 In this work, we review the basic blueprint of a computational pipeline for disease subtyping, illustrating the most common methods, choices, and challenges encountered at each stage of the analysis.…”

MATTE: a pipeline of transcriptome module alignment for anti-noise phenotype-gene-related analysis

“…To identify gene modules that correlate with the change corresponding to a different phenotype, it is critical to resolve the issue of repeatability or preservation of modules, or more specifically, whether the module will arise again in samples from various phenotypes. Differentiated modules between normal and disorder samples are usually used to identify pathogenic factors in disease studies, while preserved modules usually identify conserved and essential functions in evolution studies [ 15–19 ]. However, there are several difficulties in identifying phenotype-related gene modules.…”

Single cell analysis of hub gene characteristics of atherosclerosis based on machine learning and analysis of immune correlation of aging subtypes

“…Machine learning has a strong classi cation ability, which can help to improve the reliability and accuracy of diagnosis systems for many diseases. It has been widely used to learn the representation of features from gene expression data [7,8]. At present, the use of bioinformatics analysis and machine learning and other database veri cation is an important way to understand the pathological mechanism at the molecular level.…”