An absolute measure of similarity (e.g. Spearman correlation) between samples or genes does not in itself convey how uncommon that similarity is. Hence, in addition to computing the similarity (sim) between designated samples or genes, it is also useful to compare this similarity value to a reference distribution of similarity values (SIMnull), which can aid in interpretation of sim. To that end, we compute recall (R) as the fraction of SIMnull that is lower than sim. High R values correspond to unusually high values of sim. Thus, R provides an assessment of how well a particular pair of samples or genes match each other relative to an appropriate null.
