Preferred analysis methods for Affymetrix GeneChips revealed by a wholly defined control dataset

Area of Science:

• * Genomics
• * Bioinformatics
• * Molecular Biology

Background:

• * Increasing number of RNA-profiling data analysis methods necessitates performance evaluation.
• * Control datasets are crucial for assessing the accuracy and reliability of these methods.

Purpose of the Study:

• * To develop and utilize a novel 'spike-in' experiment for Affymetrix GeneChips.
• * To create a defined dataset for evaluating various analysis options for identifying differentially expressed genes.
• * To determine optimal analysis strategies for minimizing false-positive and false-negative rates in gene expression analysis.

Main Methods:

• * A 'spike-in' experiment was designed with 3,860 RNA species.
• * 100-200 RNAs were spiked at each fold-change level (1.2 to 4-fold) for accurate rate estimation.
• * A constant set of 2,551 RNA species was used to identify true positives and negatives.

Main Results:

• * Significant variation exists in the ability of different analysis methods to identify differentially expressed genes.
• * False-negative and false-positive rates were minimized by specific analysis choices: subtracting nonspecific signal, intensity-dependent normalization, and incorporating signal intensity-dependent standard deviation.
• * A combination of methods achieved ~70% true positive detection before a 10% false-discovery rate.

Conclusions:

• * A best-practice combination of analysis methods enhances differential gene expression detection.
• * Areas for improvement include better estimation of false-discovery rates and reduction of false-negative rates.
• * The developed spike-in experiment serves as a valuable tool for RNA-profiling data analysis method evaluation.