Quantified processed microarray data for 22 individuals with and without proliferative retinopathy
Quality control from RNA extraction was performed using the Agilent bio-analyzer, processed using the Illumina™ TotalPrep™-96 RNA Amplification Kit (ThermoFisher 4393543), hybridized to Illumina HT12v4 microarrays (Catalog number: 4393543), and scanned on an Illumina HiScan scanner  . For each of the 22 individuals, three biological replicates were profiled, with each sample being assessed at both standard glucose conditions (11mM of glucose), as well as high glucose conditions (30mM of glucose). Biological replicates were split from the same mother flask; cells were grown in separate flasks and run on different microarray plates on different days. Each biological replicate was generated from a separate frozen aliquot of that cell line. The gene expression data comprised a total of 144 samples from 22 individuals (3 replicates per individual and treatment, except for 3 individuals with 5 replicates). Gene expression was assessed in two conditions, standard glucose and high glucose, and generated from four different groups of gene expression arrays run at a given time that were carefully designed to minimize potential batch effects. BeadChip data were extracted using GenomeStudio (version GSGX 1.9.0) and the raw expression and control probe data from the four different batches were preprocessed using a lumiExpresso function in the lumi R package [8, 9] in three steps: (i) background correction (lumiB function with the bgAdjust method); (ii) variance stabilizing transformation (lumiT function with the log2 option); (iii) normalization (lumiN function with the robust spline normalization (rsn) algorithm that is a mixture of quantile and loess normalization). To remove unexpressed probes, we applied a detection filter to retain probes with strong true signal by applying Illumina BeadArrays detection p-values < 0.01 followed by removing probes that did not have annotated genes, resulting in a total of 15,591 probes.