SIG: Graph-Based Cancer Subtype Stratification With Gene Mutation Structural Information.

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Somatic tumors present high-dimensional, sparse, and small sample size data, challenging cancer subtype stratification using genomic data.
• Existing clustering methods often overlook crucial associations between mutated genes within patient-gene matrices.
• Gene mutation structural information implicitly contains cancer subtype signals, offering potential for improved clustering.

Purpose of the Study:

• To introduce a novel method, Structural Information within Graph (SIG), for enhanced cancer subtype clustering.
• To leverage gene mutation structural information to improve the accuracy of cancer patient stratification.
• To address the limitations of current methods by incorporating inter-gene mutation associations.

Main Methods:

• Constructing a graph by establishing pairwise associations between mutated genes within individual patient samples.
• Generating a structural information graph by merging these associations across all mutated genes.
• Integrating somatic tumor genomic data with the enriched gene network and applying network propagation for patient clustering.

Main Results:

• SIG effectively captures and utilizes gene mutation structural information for cancer subtype clustering.
• The method demonstrates superior clustering performance compared to state-of-the-art (SOTA) approaches.
• Validation experiments on ovarian and Lung Adenocarcinoma (LUAD) datasets confirm the efficacy of SIG.

Conclusions:

• Gene mutation structural information is a valuable, previously underutilized resource for cancer subtype stratification.
• The SIG method offers a significant advancement in cancer clustering by incorporating network-based gene associations.
• SIG has the potential to improve our understanding of cancer heterogeneity and guide personalized treatment strategies.