dualKS

dualKS performs discriminant analysis and multiclass classification of gene expression datasets using a Kolmogorov-Smirnov rank-sum algorithm to identify class-specific gene signatures.

Key Features:

• Gene Set Enrichment Analysis inversion: dualKS inverts the Gene Set Enrichment Analysis (GSEA) process to identify class-specific gene signatures.
• Kolmogorov-Smirnov approach: dualKS employs the Kolmogorov-Smirnov statistic both to define class-specific signatures and to classify samples.
• Parsimonious gene signatures: dualKS identifies highly parsimonious gene signatures; in a comparison across 10 datasets the optimal DKS signature was smaller than those of other methods in 5 of 10 cases.
• Error rate comparison with random forest: the estimated classification error using the optimal DKS signature was lower than random forest in 4 of 10 datasets and equivalent in 2 datasets.

Scientific Applications:

• Microarray studies: applied to microarray gene expression datasets where dimensionality reduction and interpretation of concise gene signatures are required.
• Multiclass classification problems: applied to multiclass genomic classification tasks to discriminate more than two phenotypic or experimental classes simultaneously.

Methodology:

dualKS ranks genes using the Kolmogorov-Smirnov rank-sum statistic to assess discriminatory power, inverts the GSEA process to define class-specific signatures, and applies the KS statistic for sample classification.