Supplementary Information: Empirical array quality weights in the analysis of microarray data

Matthew E. Ritchie¹, Dileepa Diyagama², Jody Neilson³, Ryan van Laar², Alexander Dobrovic³, Andrew Holloway² and Gordon K. Smyth¹

1. Bioinformatics Division, The Walter and Eliza Hall Institute of Medical Research.
2. Ian Potter Foundation Centre for Cancer Genomics and Predictive Medicine, The Peter MacCallum Cancer Centre.
3. Molecular Pathology Research, Department of Pathology, The Peter MacCallum Cancer Centre.

Additional Plots

• Compare false discovery rates obtained using different array weight estimates (full and gene-by-gene)
• Background intensities for arrays from the METH experiment

R Scripts

• QC data set (figures 4 a, 5 and 6): qc.R
• METH experiment (table 4 and figure 4 b): meth.R
• Simulation study - 3 arrays (table 3 and figure 2): 3arrays.R
• Simulation study - 5 arrays (table 3 and figure 3): 5arrays.R
• Simulation study - 10 arrays (figure 7): sim.R

Data

R data files used for the analyses:

• QC data set (57.1Mb)
• METH experiment (11.3Mb)
• 3 array normal simulations (2.3Mb)
• 3 array empirical simulations (2.3Mb)
• 5 array normal simulations (2.4Mb)
• 5 array empirical simulations (2.4Mb)

Required Software

• limma: http://bioinf.wehi.edu.au/limma/
• R: http://cran.r-project.org/