MAQC

The Microarray Quality Control project provides tools and data to the community so that gene expression platforms can be benchmarked and compared to one another. This dataset consists of RNA-Seq libraries made from “off-the-shelf” RNA obtained from two sources: the Stratagene Universal Human Reference, a mixture of RNA from 10 cell lines, and Ambion human brain RNA. Seven replicates of each were sequenced, and these were grouped in various ways to assess the empirical false positive rates of Cuffdiff and other methods. The table below describes the contrasts performed:

Contrast	Group A	Group B	Description
brain_vs_uhr	Brain 1-7	UHR 1-7	Should produce many differentially expressed genes, transcripts, and genes with various types of isoform switching
brain_only_A	Brain 1-3	Brain 4-6	Should produce no differentially expressed genes. Since each group is from the same condition, the variability within groups is extremely low.
brain_only_B	Brain 1-2	Brain 4-5	Should produce no differentially expressed genes. Since each group is from the same condition, the variability within groups is extremely low.
brain_only_C	Brain 1	Brain 4	Should produce no differentially expressed genes. Since each group is from the same condition, the variability within groups is extremely low.
nonsense_A	Brain 1,3,5, UHR 2,4,6	Brain 2,4,6 UHR 1,3,5	Should produce no differentially expressed genes. Since each group contains a mixture from both Brain and UHR, the variability within groups is extremely high.
nonsense_B	Brain 4,5,7, UHR 4,6,2	Brain 1,6,2, UHR 3,1,5	Should produce no differentially expressed genes. Since each group contains a mixture from both Brain and UHR, the variability within groups is extremely high.

The above plot shows, for each of the MAQC contrasts, the number of significantly differentially expressed genes and transcripts, as well as various types of isoform switching reported by Cuffdiff 2. Significant DE genes reported by DESeq and edgeR are included for comparison.