Data sets were selected for this study based on the following criteria: (1) chip platform (only data from Affymetrix Human Genome U133 Plus 2.0 expression arrays was included); (2) human primary cell; (3) cell-subset studied; (4) availability of raw data (.cel) and sample description files. Quality control (QC) of these data using the arrayQualityMetrics package in Bioconductor showed a number of chips/data sets to be of poor quality or not comparable based on the chip signal intensities. Furthermore, additional datasets were rejected from this analysis following initial analysis, as these data showed inexplicable differences in their global expression intensities when compared to data from supposedly similar cells. Out of the 1,103 chips originally selected from 105 separate studies meeting the above criteria, 745 arrays passed the criteria for further analysis on the basis of these QC arrays. All data was then subjected to RMA normalisation.
