pvac

pvac filters probe-level gene expression data using principal component analysis to remove non-informative genes and improve detection of differentially expressed genes in large-scale array studies such as Affymetrix GeneChips®.

Key Features:

• Principal Component Analysis: Applies PCA to probe-level gene expression data and quantifies the proportion of variation captured by the first principal component (PVAC).
• Filtering Strategy: Uses PVAC to filter genes, reducing the number of null hypotheses tested across tens of thousands of genes in array experiments.
• Improved Sensitivity and Specificity: Increases sensitivity for detecting differentially expressed genes while controlling false discovery rates by excluding genes that contribute little to the first principal component.
• Data-Driven Threshold Selection: Selects the PVAC-based filtering threshold via a data-driven procedure that determines an optimal cutoff based on dataset characteristics.

Scientific Applications:

• Clinical biomarker discovery: Identifying diagnostic and prognostic biomarkers from gene expression arrays in clinical studies.

Methodology:

Performs PCA on probe-level expression data to compute the proportion of variation explained by the first principal component (PVAC), applies a PVAC-based filtering threshold selected via a data-driven procedure, and reduces the gene set for downstream differential expression testing.