sdef

sdef identifies features that are significant across multiple microarray experiments to detect genes consistently differentially expressed.

Key Features:

• Statistical Methods: Implements a permutation test that assesses the maximal ratio of observed to expected common features under the hypothesis of experiment independence and a Bayesian framework that accounts for uncertainty in the number of differentially expressed genes.
• Input Measures: Operates on ranked lists of p-values or other significance measures to evaluate overlap between experiments.
• Application to Public Datasets: Applied to microarray datasets including a Type 2 diabetes susceptibility study in mice (liver and skeletal muscle) and cross-mammalian sex comparison experiments.

Scientific Applications:

• Type 2 diabetes (mouse tissues): Identified between 68 and 104 genes commonly perturbed across liver and skeletal muscle and detected enrichment of inflammation pathways relevant to obesity and diabetes.
• Mammalian sex comparisons: Identified 110 genes commonly affected across three experiments and noted enrichment in genes associated with cell development.

Methodology:

Permutation testing of observed-versus-expected overlaps under independence; Bayesian modeling incorporating priors to account for uncertainty in the number of differentially expressed genes; analysis of ranked lists of p-values or other significance measures.