Multi-omics subtyping of hepatocellular carcinoma patients using a Bayesian network mixture model

Area of Science:

• Computational biology
• Bioinformatics
• Cancer genomics

Background:

• Accurate cancer subtyping is crucial for personalized medicine.
• Multi-omics data integration offers deeper insights into cancer heterogeneity.

Purpose of the Study:

• To introduce bnClustOmics, a statistical model and computational tool for unsupervised multi-omics clustering.
• To characterize cancer subtypes by learning networks of omics variables within clusters.
• To identify potential therapeutic targets through molecular subgroup analysis.

Main Methods:

• Bayesian network mixture model for patient sample clustering.
• Unsupervised learning of omics variable networks for each cluster.
• Application to a hepatocellular carcinoma (HCC) multi-omics dataset (genome, transcriptome, proteome, phosphoproteome).

Main Results:

• Identified three distinct hepatocellular carcinoma (HCC) subtypes.
• Characterized molecular features of each HCC subtype.
• Discovered subtype-specific networks linking genotypes to molecular phenotypes.
• Found molecular characteristics associated with patient survival, independent of clinical stage.

Conclusions:

• bnClustOmics effectively clusters multi-omics data and reveals molecular characteristics of cancer subtypes.
• The discovered networks provide a deeper understanding of cancer biology and genotype-phenotype relationships.
• This approach facilitates the identification of personalized treatment strategies for cancer patients.