metaArray

metaArray performs probabilistic meta-analysis of microarray gene expression data to integrate high-throughput datasets across multiple studies and identify consistent expression and differential-expression signals.

Key Features:

• Probabilistic framework: Uses a general probabilistic framework with mixture models to combine high-throughput genomic data from multiple related microarray experiments.
• Latent variables: Incorporates latent variables that represent quantities combinable across different experimental setups and platforms.
• Probability of Expression (POE): Estimates the Probability of Expression (POE) index to quantify the likelihood of gene expression events across studies.
• Estimation methods: Provides POE estimation via Markov Chain Monte Carlo (MCMC) and an Expectation-Maximization (EM) algorithm.
• Signed probability transformation: Transforms gene expression data into a signed probability scale using MCMC/EM-based estimations.
• Combined differential expression: Performs combined differential-expression analysis on the raw scale by applying a weighted Z-score after stabilizing the mean–variance relationship within each platform.

Scientific Applications:

• Meta-analysis of disease datasets: Integration of microarray studies to identify consistent expression patterns across experiments, including datasets related to metastatic cancer.
• Biomarker and target identification: Identification of consistent biomarkers and therapeutic targets through cross-study differential-expression analysis.

Methodology:

Employs mixture models with latent variables to estimate the Probability of Expression (POE) using MCMC or EM, transforms expression values to a signed probability scale, stabilizes mean–variance relationships within each platform, and combines results via a weighted Z-score for differential-expression analysis.