Figure 2
From: Revealing protein networks and gene-drug connectivity in cancer from direct information
Protein-protein interaction pairs predicted from pan cancer protein expression levels with DCA. (a) Positive prediction values comparison between the direct information and mutual information. Solid line plots indicate number of protein-protein pairs verified with a curated literature review and STRING pairs with a combined scored at medium confidence (y-axis) against total number of pairs (x-axis) ranked by DI values (red) or MI (blue) values. Dash line plots indicate coupling pairs identified as binary interactions classified from a literature review and STRING pairs with experimental scores at highest confidence. The ratios of direct interactions to the all verified interactions for top 100 pairs are shown in the inset table. (b) Comparison of top 100 DI ranked and MI ranked predicted pairs. The pie charts on the right panel represent the distinct protein pairs in both groups with different colors indicating different categories. The digits near the pie charts are the number of protein pairs included into according category. Two lists named Ranked Top 25 show the detailed protein pairs that appear distinctly in the top 25 pairs by DI and MI methods.
