Antibody-microarray data preparation
To determine spot intensities, we calculated the mean pixel intensity per spot. To determine background intensities we calculated the median pixel intensity per background “doughnut” (Additional file 1: Figure S1D). Individual array spots were background subtracted locally (by subtracting the median background across spot replicates in each sample). Spots with a residual intensity less than 10 % above background were set to ‘ND’ (non-detectable). Antibodies with more than 55 % ‘ND’ values were excluded from the analysis (N = 11), yielding a total of 582 quantifiable antibodies (Additional file 2, for ‘ND’-count distribution see Additional file 1: Figure S2C). ‘ND’ values were then replaced with the greater of the half the minimum non-‘ND’ value per sample replicates, the half the minimum non-’ND’ value of that antibody across all samples (if the sample replicates were all ‘ND’), or 1. The spot data were Log2 transformed, replicate averaged, and iteratively (i = 50) row- and column-wise median centered (subtract the column-wise median from the values in each column/row of data, so that the mean or median value of each column/row is 0) and normalized (multiply all values in each column/row of data by a scale factor S so that the sum of the squares of the values in each column/row is constant across columns/rows) following a procedure described in the Cluster 3.0 manual [63]. Finally the data were Z-scored, leaving approximately normally distributed data for analysis with a mean of 0 and a standard deviation of 1 (Additional file 1: Figure S2D to F, 86 % of all antibodies have normal distributions based on one-sample Kolmogorov-Smirnov test).