Network based stratification of major cancers by integrating somatic mutation and gene expression data

Area of Science:

• Genomics
• Bioinformatics
• Cancer Research

Background:

• Cancer subtyping is crucial for targeted prognosis and treatment.
• Existing methods like network-based stratification (NBS) may overlook cancer-specific heterogeneity.
• Integrating diverse molecular data can enhance subtype discovery.

Purpose of the Study:

• To develop and validate a novel cancer subtyping method using cancer-type-specific co-expression networks and mutation data.
• To improve the precision of patient clustering for better clinical outcome association.
• To compare the proposed method against the established NBS approach.

Main Methods:

• Construction of cancer-type-specific significant co-expression networks (SCNs) from gene expression data.
• Propagation of somatic mutation data onto SCNs for clustering.
• Application of an improved network-regularized non-negative matrix factorization (netNMF_HC) for precise classification.
• Validation using The Cancer Genome Atlas (TCGA) datasets for ovarian cancer (OV), lung adenocarcinoma (LUAD), and uterine corpus endometrial carcinoma (UCEC).

Main Results:

• The proposed method successfully identified survival-relevant cancer subtypes across multiple cancer types.
• The algorithm demonstrated superior performance compared to the NBS method in identifying informative subtypes.
• Specifically, novel survival-associated subtypes were identified in UCEC that were missed by NBS.
• The method effectively leverages mutation data and cancer-specific co-expression patterns for precise subtyping.

Conclusions:

• Cancer-type-specific SCNs combined with mutation data and netNMF_HC offer a powerful approach for patient subtyping.
• This method enhances the identification of clinically relevant cancer subtypes, improving prognostic and therapeutic strategies.
• The findings underscore the importance of accounting for cancer heterogeneity in subtyping algorithms.