This paper evaluates the reproducibility of microarray technology and presents metrics for evaluating quality and reproducibility. The authors use data from the Agilent microarray system and Rosetta Resolver algorithms to demonstrate that the system produces data with high confidence for gene expression interpretations. The paper also provides metrics for calculating the reproducibility of microarray technology, including metrics for the microarray preparation process, signal statistics for each gene on each array, signal statistics for each gene across microarrays spotted with the same sample, log ratio statistics for all genes on each self-self microarray, log ratio statistics for each gene on each differential expression microarray, log ratio statistics for each gene across all differential expression microarrays, and biological replicate metrics.
This document serves as a guide for evaluating the reproducibility of microarray technology. It details experimental designs, such as self-self and dye-swap arrays, required to assess data quality and confidence in differential expression results.
The paper presents comparisons of various metrics with and without features identified as anomalous by Agilent Feature Extraction software. These metrics include signal statistics (%CV) for each gene across microarrays and log ratio statistics (SD), demonstrating how high reproducibility increases confidence in gene expression interpretations.
Researchers need to design four self-self microarrays and four dye-swap differential expression microarrays.
Each microarray should contain at least 10 genes, and each gene should have at least 15 feature replicates.
The software flags features if they are non-uniform in either the green or red channel, or saturated in both channels.