Powers of multiple-testing procedures for identification of genes significantly differentially expressed in...

Area of Science:

• Genomics
• Bioinformatics
• Statistical Genetics

Background:

• High operational costs of microarray experiments necessitate efficient experimental design.
• Accurate determination of replicate numbers is vital for detecting significant gene expression changes.

Purpose of the Study:

• To provide guidance for microarray experimental design by calculating the power and replicate numbers required for multiple-testing procedures.
• To evaluate the performance of different multiple-testing procedures in detecting differential gene expression.

Main Methods:

• A mixture distribution model was applied to microarray data.
• A single-distribution model of t-statistic was fitted to a rat stroke dataset (3,000 genes).
• Power calculations were performed for four popular multiple-testing procedures to detect differential gene expression.

Main Results:

• The Benjamini-Hochberg (BH)-procedure demonstrated the highest power in detecting small gene expression changes compared to other procedures.
• All tested multiple-testing procedures showed equal power in identifying genes with the largest expression changes.
• The power of the BH-procedure to detect small changes remained consistent regardless of the total number of genes, unlike other procedures.

Conclusions:

• The choice of multiple-testing procedure significantly impacts the power to detect differential gene expression in microarray studies.
• The BH-procedure is recommended for its robust performance in identifying small expression changes, contributing to optimized experimental design and cost savings.