CoINcIDE: A framework for discovery of patient subtypes across multiple datasets

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Patient disease subtypes hold promise for personalized medicine.
• Many existing subtype discovery methods lack cross-dataset reproducibility, limiting clinical application.
• High-dimensional omics data presents challenges for robust subtype identification.

Purpose of the Study:

• To introduce CoINcIDE, a novel computational framework for discovering reproducible patient subtypes across multiple datasets.
• To present curatedBreastData, a comprehensive database of over 2,500 breast cancer gene expression samples.
• To demonstrate the utility of CoINcIDE in identifying novel, prognostically significant cancer subtypes and potential therapeutic targets.

Main Methods:

• Developed CoINcIDE, a methodological framework that does not require between-dataset transformations for subtype discovery.
• Curated and compiled the curatedBreastData database from multiple gene expression datasets.
• Applied CoINcIDE to breast and ovarian cancer datasets to identify and validate novel subtypes.

Main Results:

• CoINcIDE successfully identified novel breast and ovarian cancer subtypes with significant prognostic value across multiple datasets.
• The framework demonstrated robustness and reproducibility without the need for data normalization between datasets.
• Hypothesized novel therapeutic targets for ovarian cancer were identified based on the discovered subtypes.

Conclusions:

• CoINcIDE offers a reliable approach for patient subtype discovery, overcoming limitations of previous methods.
• The identified subtypes and therapeutic targets have potential implications for advancing personalized cancer medicine.
• CoINcIDE and curatedBreastData are publicly available as R packages to facilitate further research.