Identifying cancer sub-types from genomic scale data sets using confidence based integration (CBI).

Area of Science:

• Genomics
• Bioinformatics
• Precision Medicine

Background:

• Precision medicine aims for refined patient diagnosis by identifying underlying causes in similar patient cohorts.
• Advancements in technology enable feature extraction from diverse sources like genetics, generating vast datasets for micro-studies.
• Integrative methods using genomic datasets are crucial for nuanced cancer research, leveraging resources like The Cancer Genome Atlas (TCGA).

Purpose of the Study:

• To introduce Confidence Based Integration (CBI), a novel method for integrating diverse genomic datasets.
• To extract similar and complementary information from gene expression, miRNA, and DNA methylation data.
• To improve patient stratification and understanding of disease subtypes and their survival prospects.

Main Methods:

• Applied Confidence Based Integration (CBI) to fuse gene expression, miRNA, and DNA methylation datasets from five cancer types.
• Utilized Spectral Clustering algorithm on fused data to derive distinct disease subtypes.
• Evaluated cluster significance using the log-rank test for survival analysis.

Main Results:

• CBI successfully integrated multi-omics data, revealing distinct cancer subtypes with significant survival differences.
• The fused data clusters demonstrated higher biological relevance compared to individual data features.
• CBI significantly outperformed Similarity Network Fusion (SNF) and other methods in forming statistically significant clusters (lower P-values in log-rank tests).

Conclusions:

• Confidence Based Integration (CBI) is a robust method for multi-omics data fusion in cancer research.
• CBI enhances the identification of biologically meaningful cancer subtypes, improving patient stratification.
• The approach provides valuable insights into patient prognosis and the molecular underpinnings of disease heterogeneity.