Skip to main content
. 2014 Nov 26;31(7):1093–1101. doi: 10.1093/bioinformatics/btu786

Fig. 3.

Fig. 3.

Our multiple human interaction networks allow improved human pathway component retrieval. In panel A, we detail a combination of transcriptional, post-translational and physical interactions that we accurately prioritize surrounding the human tumor suppressor FOXO3. In panel B, we expand our pathway retrieval evaluation analysis to total 447 human curated pathways. Each dot in the box plot represents our accuracy of ranking the pathway interactions above all incorrect interactions between the constitute genes in a pathway. Overall, our predicted networks show significantly improved performance in retrieval of pathway interactions when evaluated for both all interactions types combined (‘all labels’) and also for each individual interaction type (shown in salmon color) compared with average correlation over the expression dataset (‘all labels (Average correlation)’ in lime-green)