Cancer subtype identification using somatic mutation data

Area of Science:

• Genomics
• Cancer Biology
• Bioinformatics

Background:

• Next-generation sequencing has advanced cancer driver mutation identification.
• Large-scale cancer sample screening generates abundant data for analysis.
• Patient classification by mutation profiles aids in identifying treatment subgroups, but data sparseness and heterogeneity pose challenges.

Purpose of the Study:

• To introduce a novel method for de-sparsifying somatic mutation data.
• To apply this method to a large dataset across multiple cancer types.
• To identify patient subtypes and potential therapeutic associations.

Main Methods:

• Developed a pathway-based de-sparsification method for somatic mutation data.
• Applied the method to 5805 primary tumor samples from 23 cancer types in The Cancer Genome Atlas.
• Integrated phenotypic data for association analysis.

Main Results:

• De-sparsified mutation data showed associations with phenotypic data across most cancer types.
• Identified poor prognostic subtypes in three cancer types linked to signal transduction pathways with available targeted treatments.
• Discovered subtype-drug associations for 14 additional subtypes.
• Uncovered nine pan-cancer subtypes associated with four overarching biological pathway sets.

Conclusions:

• The study represents a significant advancement in understanding cancer mutational patterns.
• The developed method facilitates the identification of clinically relevant cancer subtypes.
• Findings support the potential for pathway-informed precision oncology.