When selecting SNR as the metric in the GSEA Workbench in Qlucore Omics Explorer the user also specifies a sample annotation and a group in that annotation. If we call the selected group A and put the remaining sample groups into a group common group we call B, SNR is then computed for each variable x as
SNR(x) = (mean(x in A) - mean(x in B)) / (std(x in A) + std(x in B))
where std is the standard deviation.
So essentially the "signal" is the difference in mean value of the measured variable between between the two groups. While the "noise" is the sum of the variation of the measured variable in the two groups. Thus if the group variability is high for a given variable, the signal (group mean value difference) needs to be high in order not to drown in the random variations of the variable over the samples. |
