Figure 3

Estimating false-positive rates of large-scale assays. (a) As described by D'haeseleer and Church [32], the number of true positives in an interaction dataset can be estimated by examining the ratio of intersections of two similar datasets (A and B) and a reference dataset. If intersections contain all true positives, then the ratio of areas I and II is equal to the ratio of areas III and IV, where IV contains true positives (and V false positives, not shown to scale). The number of false positives can then be determined by simple subtraction, repeating the calculation for the other dataset. (b) Calculation of false-positive rates for the most recent yeast mass spectrometry assays of Gavin et al. [27] and Krogan et al. [28] within the interactome subspace sampled by both experiments (1,243 baits) and using MIPS as the reference sample [33]. Intersections (regions I, II, II) were determined by examining the data, and true- and false-positive populations (regions IV and V) were calculated as described in (a).