By tissue sample comparison we were able to identify nine common tissues represented in both experiments (Frontal cortex, heart, kidney, liver, lung, pancreas, spleen, stomach and thyroid).
We performed several descriptive sample comparisons: We did a drop-out analysis by comparing the sample of 3,220 patients finally included in statistical analyses with the sample excluded from analyses due to missing data on AMDP (N = 2,004) in terms of basic admission-specific patient characteristics (see table 1 also for applied statistical tests).
For inter-sample comparison, we normalized these expression levels by dividing the values by the slide median, i.e. the median of all probes on the array.
Because we wished to estimate the expected performance of the model for species that were not used for fitting and to avoid the unrealistically low error rates returned by in-sample comparisons, we randomly divided the data, stratifying by class, into training (75%), and test (25%) sets (22,23).
To identify differentially activated TFs for two-sample comparisons, we propose incorporating two tests measuring different aspects of the conditional relevance of a TF.
For cross-sample comparisons, we normalized raw read counts for these samples using the TMM method (trimmed-mean of M values; a weighted trimmed mean of the log expression ratios [ 53]) as implemented in the edgeR package [ 54] (http://www.bioconductor.org).
Because CNV-seq requires a control sample for comparison, we performed all possible pair-wise comparisons of the 5 genomes.
As we were only able to obtain two reference samples for comparison, we cannot be certain whether most failed samples were counterfeit or substandard.
In order to have enough normal tissue samples for comparison, we obtained additional normal tissue slides from Tissue Procurment Core at Washington University in St . LouisSchool of Medicine according to the approved protocol by Washington University in St Louis Human Studies Committee, such that the total number of normal tissues was 65.
For example, based on morphological (length variation) and genetic (pairwise FST using only the three candidate SNPs from the 'Bonneville' sample analyses) comparisons, we observed some significant differences between the sample of translocated fish, and the two samples of randomly selected fish from different time periods in the migration season.
We used a typical number of arrays for a given sample comparison in a large scale screening experiment.
