Figure 5

Visualization of the empirical log-p-values from ALL dataset for the top-100 classes of each scoring function. Part A shows the obtained p-values when all class randomizations are normalized and pooled. Part B shows the results when each class is analyzed separately. Largest value in lower plot refers to p-value = 0 (see main text for details). Lines are: blue with circles = GSZ-score; green with cross = t-test; red with squares = KS; magenta with x = iGA; cyan with diamonds = modKS. Notice that GSZ-score shows here a very clear separation from the other methods in both of the plots. Any reasonable threshold would result in a larger number of significant classes with GSZ-score than with any other method.