Transcriptomic pan-cancer analysis using rank-based Bayesian inference.

Area of Science:

• Genomics
• Bioinformatics
• Cancer Research

Background:

• Analyzing large pan-cancer datasets for robust subgroup identification is challenging.
• Existing methods may lack comprehensive uncertainty quantification and clear biological interpretability.

Purpose of the Study:

• To develop and apply a novel rank-based Bayesian clustering method for pan-cancer data analysis.
• To identify robust, biologically interpretable subgroups within diverse cancer types.

Main Methods:

• Applied a novel rank-based Bayesian clustering approach to RNA-seq data from 12 Cancer Genome Atlas tumor types.
• Integrated and quantified uncertainties from input data and the model.
• Characterized clusters by top-ranked genomic features for biological interpretation.

Main Results:

• Identified a robust clustering reflecting both tissue of origin and pan-cancer relationships.
• Discovered three pan-squamous clusters (lung, head/neck, bladder) with distinct biological functions.
• Uncovered two novel kidney cancer subtypes with differential prognoses and validated known breast cancer subtypes.

Conclusions:

• The developed Bayesian clustering method effectively identifies robust and biologically meaningful clusters in pan-cancer samples.
• The method provides probabilistic outputs for assessing cluster stability and reliable biological interpretation.
• Findings highlight potential for discovering new cancer subtypes and understanding pan-cancer biology.