Peculiar Genes Selection (PGS)

Peculiar Genes Selection (PGS) identifies class-specific biomarker genes from high-dimensional genomics and transcriptomics data to improve classification performance, particularly in imbalanced datasets.

Key Features:

• Three-Step Feature Selection Methodology: Performs differential expression detection, filters features by discriminative power, and identifies class-specific features that form biomarker signatures.
• Handling Imbalanced Data: Targets datasets with imbalanced class sizes and has been demonstrated to improve classification performance using Support Vector Machines (SVM), achieving 82% accuracy versus ~73% for comparative methods.
• Performance on Balanced Datasets: Shows comparable performance to existing methods when applied to perfectly balanced datasets.
• Biological Relevance Validation: Validates selected biomarker signatures using Gene Ontology enrichment analyses to confirm biological relevance.

Scientific Applications:

• Genomics and Transcriptomics: Identification of class-specific biomarkers in high-dimensional genomics and transcriptomics datasets.
• Disease Classification: Selection of discriminative gene signatures to support disease class prediction.
• Personalized Medicine: Enabling more accurate patient stratification based on selected biomarker signatures.
• Biomarker Discovery: Facilitating discovery and validation of biomarkers for downstream biological interpretation and functional analysis.

Methodology:

Identifies differentially expressed genes, filters features by discriminative power, detects class-specific features, performs classification with Support Vector Machines (SVM), and evaluates biological relevance via Gene Ontology enrichment.