Expression values are z-scores across genes in each sample of a dataset. Directly comparing expression patterns across datasets can be misleading due to the varying biases associated with the total number of observed genes and the observations across different techniques.
Co-Expression Network
This gene co-expression network is constructed based on all datasets by Fisher’s combination of corrected Spearman correlation p-values. Specific thresholds are used for generating the whole network. The genes are restricted to protein network data (incl. distinction: direct vs. interologs) in STRING. A two-step network is extracted from the whole network. This two-step network consists of genes directly linked with the query gene (first-order links) as well as genes linked to those directly connected genes (second-order links).