Functional Characterization of Transcription Factor Motifs Using Cross-species Comparison across Large Evolutionary Distances
(A) Two examples of how different motif scanning methods were compared, corresponding to two different motifs (“CG33980” and “KNIRPS”). The y-axis plots the numbers of associations between the motif and the real promoter set, at different levels of significance (always with p-value<0.05), and the x-axis shows the number of associations with the shuffled promoter set at the same level of significance. In both examples, there is a method that is superior by all three measures used for comparison: “strong criterion”, “AUC” and “N0” (see Methods). For KNIRPS, while site-LLR performs best, Stubb dominates SWAN by the “AUC” measure, SWAN dominates Stubb by the “N0” measure (the value of y at x = 0), and there is ambiguity in terms of the “strong criterion”. (B) Comparison of different motif scanning methods, using the number of motifs for which each method performed best as per the AUC criterion. Left panel: comparison of site-LLR, SWAN, and Stubb. Right panel: evaluating the effect of “PGC” parameter (see Methods).