Figure 2

Evidence supporting that experimental sampling of protein interaction networks is not random. The ratio is defined as the proteins which link to 10 or more proteins when used as baits but none as preys versus either the total proteins of the network (magenta) or the total proteins who link to more than 10 proteins as bait no matter how many proteins linked to when used as preys (blue). a. Experimental datasets of yeast, Drosophila, and human. b. Truncated power-law network for different interaction retention rates. The sampling was done as follows. We first generated a large network and random sampled the proteins and interactions in a way that matches the size and number of links in the Drosophila dataset. Note that to maintain the sampled network at the size of the Drosophila dataset for different retention rates of interactions, one needs to start with networks of different original sizes. 5000 random samplings were carried out for each retention rate. Bars represent the average ratio and triangles are the maximum of the 5000 samplings.