CORAZON

CORAZON performs normalization and unsupervised clustering of gene expression datasets to identify expression-based gene groups for functional characterization and downstream genomic analyses.

Key Features:

• Unsupervised clustering: Implements Mean Shift, K-Means, and Hierarchical Clustering to group genes by expression profiles.
• Normalization and transformation: Provides eight normalization/transformation methods applicable to gene expression data.
• Data type support: Supports RNA-seq, microarray, and NanoString gene expression datasets.
• Gene-length-specific normalizations: Includes normalization options that require only gene length information, suitable for RNA-seq analysis.
• Cross-tissue analysis: Groups genes based on expression profiles across different tissues.
• Performance evaluation: Clustering methods were evaluated using five models with reported accuracies between 92% and 100%.
• Functional interpretation: Enables cluster-level functional characterization via Gene Ontology enrichment analyses.

Scientific Applications:

• ncRNA functional categorization (ENCODE): Analysis of non-coding RNAs from the ENCODE project revealed GO enrichment in Detection of Stimulus, Sensory Perception, Immunological System, and Digestion.
• Coding gene functional clustering: Coding genes were clustered into categories such as Cellular, Metabolic, Transports, and Systems Development.
• Cross-platform gene expression analysis: Applicable to comparative analyses across RNA-seq, microarray, and NanoString datasets.
• Downstream genomic analyses: Identification of expression-based gene groups for subsequent functional annotation and downstream analyses.

Methodology:

Applies eight normalization/transformation methods (some requiring only gene length), then performs unsupervised clustering with Mean Shift, K-Means, and Hierarchical Clustering; clustering performance was assessed on five models (92%–100% accuracy) and clusters were interpreted via Gene Ontology enrichment.