The different quality measures were highly correlated. The data show a correlation between probe-length and variability, with longer probes giving less variance in fluorescence intensity. For the SNP platforms, we observed that besides sample-specific variability, systematic effects between a sample and the reference can also greatly inflate per-chip variability estimates, and consequently the ability to make reliable CNV calls. Specifically for Affymetrix results, we found large differences in quality control values depending on the baseline reference library used (Supplementary Fig. 2). Based on these results, subsequent analysis of Affymetrix data from The Centre for Applied Genomics (TCAG) was based on an internal reference library, whereas analysis of Affymetrix 6.0 data produced at the Wellcome Trust Sanger Institute (WTSI) was done using the Affymetrix-supplied reference library (no internal reference was available).
