Spike-in validation of an Illumina-specific variance-stabilizing transformation.

Abstract

BACKGROUND: Variance-stabilizing techniques have been used for some time in the analysis of gene expression microarray data. A new adaptation, the variance-stabilizing transformation (VST), has recently been developed to take advantage of the unique features of Illumina BeadArrays. VST has been shown to perform well in comparison with the widely-used approach of taking a log2 transformation, but has not been validated on a spike-in experiment. We apply VST to the data from a recently published spike-in experiment and compare it both to a regular log2 analysis and a recently recommended analysis that can be applied if all raw data are available. FINDINGS: VST provides more power to detect differentially expressed genes than a log2 transformation. However, the gain in power is roughly the same as utilizing the raw data from an experiment and weighting observations accordingly. VST is still advantageous when large changes in expression are anticipated, while a weighted log2 approach performs better for smaller changes. CONCLUSION: VST can be recommended for summarized Illumina data regardless of which Illumina pre-processing options have been used. However, using the raw data is still encouraged whenever possible.