The PBS is computed as an e-value which varies between 0 and 1 and gives the probability for the interaction to be non specific (the same way a BLAST analysis is computed). We attributed thresholds to the computed e-values to define 4 categories ranging from A to D:

 •    PBS A, B, C characterize high confidence interactions, interactions supported by several independent fragments and/or involved in reciprocal interactions or cycles.

 •    PBS D is assigned to interactions identified through one unique fragment or multiple identical ones. They can be interactions hardly detectable by the Y2H technique (weak or transient interactions, low abundance of the transcript in the library) or represent false-positive technical background.

 •    The PBS E is a special category and represents interactions involving prey domains connected at least 10 times in screens with Human, Mouse, Drosophila and Arabidopsis cDNA libraries and 6 times for all other organisms. This arbitrary threshold allows us to flag highly connected (non-specific) protein domains. Our experience with more than 4,500 screens performed so far has allowed us to classify highly connected proteins in different categories: i) Proteins that are known to be highly connected due to their biological function (modifying enzyme, chaperones, protein degradation enzymes…) and which therefore easily exceed the threshold of 6 when many screens are performed within the same organism; ii) Proteins with a prey interacting domain that contains a known protein interaction motif (e.g. PDZ, Ras Association Domain, …) or a biochemically promiscuous motif (e.g. highly charged residues). Their higher connectivity may be specific or the consequence of the biochemical promiscuity (interaction biochemically relevant but not necessarily occurring in a physiological setting).

  •    The PBS F flags experimentally proven artifacts of the Y2H technology. These can be LexA or Gal4 protein binders, binders of the DNA sequence upstream of the reporter gene…