Figure 5

Profile pairs preferred by the unweighted hypergeometric metric without runs. Shown are the top 10 pairs of profiles that score highly in unweighted hypergeometric p-value but poorly in the runs-informed metric as determined by smallest ratios of unweighted hypergeometric p-value without runs to our runs-using score (taking both on a linear and not on a logarithmic scale). Not surprisingly, the matches between these profiles are concentrated in few runs. We find that the protein pairs here are not closely related functionally according to our snapshot of GO, and so they are likely false positives for the runs-oblivious unweighted hypergeometric model.