Yeast two-hybrid array screens and vectors. Shown are two Y2H screens with four different vector combinations. Each interaction is represented by two colonies to ensure reproducibility. (A) Lambda bait protein A (DNA packaging protein) was fused to an N-terminal DNA-binding domain ("DBD", in pGBKT7g) and was tested against prey constructs in both N- and C-terminal configurations (activation domains in pGADT7g, and pGADCg). (B) The C-terminal DBD fusion (in pGBKCg) as tested against prey constructs in both N- and C-terminal configurations (in pGADT7g, and pGADCg). The interactions of C-terminal preys are labeled with an asterisk (*), all remaining interactions use N-terminal fusions. All the interactions obtained from the array screening were subjected to Y2H retests: we were able to retest all the interactions shown in Figure 2 except A-Ea47, which has thus been removed from the final interaction list. Technical details of the screening procedure have been described in [8, 10]. (C) Interaction quality assesment. Using the experimental derived false positive rate from  and Bayes theorem, we estimated the probability of an interaction to be true. This estimate depends on the vector system, being highest (83%) for pDEST22/32, and lowest (40%) for pGBKCg/pGADT7g. (D) Detection of known PPIs with different vector systems. Known PPIs are enriched in the subset of PPIs detected by > = 2 vector systems compared to PPIs detected by 1 vector combination.