vsn

vsn normalizes microarray intensity data for single- and multiple-color arrays by applying a variance-stabilizing transformation and robust calibration to enable reliable differential expression analysis.

Key Features:

• Statistical Model for Gene Expression Data: Integrates data calibration, differential expression quantification, and measurement error assessment in a unified statistical framework.
• Variance Stabilizing Transformation (VST): Uses the transformation h(x) = arsinh(a + b x) derived from modeling the variance–mean relationship to stabilize variance across a wide intensity range.
• Difference Statistic: Computes a difference statistic Δh with approximately constant variance across the intensity spectrum, which approaches log-ratio behavior at high intensities.
• Robust Parameter Estimation: Estimates transformation parameters and between-experiment calibration using a robust variant of maximum-likelihood estimation to mitigate outliers and noise.

Scientific Applications:

• Data Pre-processing: Provides normalization and calibration for microarray pre-processing to support downstream multivariate and statistical analyses.
• Cross-platform Compatibility: Applicable to single- and multiple-color microarrays, including two-color cDNA arrays and Affymetrix oligonucleotide arrays.

Methodology:

Models the variance–mean relationship to derive the VST h(x) = arsinh(a + b x), computes the difference statistic Δh, and estimates transformation and calibration parameters via a robust variant of maximum-likelihood estimation while assessing measurement error and differential expression.