Skip to main content
Fig. 8 | BMC Bioinformatics

Fig. 8

From: Simultaneous learning of individual microRNA-gene interactions and regulatory comodules

Fig. 8

ARI and F\(^{1}\) score comparison between Theia, SNMNMF, PIMiM, and Tiresias. The ARI and F\(^{1}\) score of SNMNMF and PIMiM are compared relative to Theia, i.e., the more negative the value, the better Theia performed relative to the alternative methods. A value greater than zero indicates that Theia was outperformed and zero indicates that both methods performed equally. In each case, the x-axis indicates the regulation strength and the y-axis indicates the relative false positive rate in the putative miRNA-gene interactions. Note that the x-axis stops at 0.15 for the F\(^{1}\) comparison because all of the methods except Tiresias reached their maximum potential score at or before this point. The ARI for Tiresias cannot be calculated because Tiresias only predicts miRNA-gene interactions; thus, we plot the F\(^{1}\) score a second time (bottom left) with a different axis to show Tiresias reaching its maximum potential F\(^{1}\) score. In this plot, we see that Tiresias has a slight advantage over Theia in recovering interactions when the regulation strength is high. However, at a low regulation strength, which is biologically more relevant, Theia significantly outperforms Tiresias. We can also see that Theia achieves a significantly higher ARI than SNMNMF and PIMiM

Back to article page