Referee 3's review:

This paper describes a method for the functional characterization of proteins based on domain-domain interactions and guilt-by-association.

The work is somewhat interesting but the authors fail to cite relevant work so that the readers are confused as to the actual impact of the present study.

MAJOR POINTS

1. The authors claim that their method is reliable and "practical to use" how does this connect with the software availability.

2. There are similar pieces of work published recently, which the authors do not cite. They will probably need to cite the following PMIDs: 17647236 / 17456608 / 16725050 and even less recently: 11473021. The authors will need to explain eloquently how their method differs from those.

3. The functional distance is explained in some detail in Methods, but does not cite work that has been reported previously, e.g. PMID 17344234 / 16776819, 12835272. These need to be cited.

4. The clustering procedure is a bit vague, sounds like K-means clustering. Can the authors specify and map it to something that is more well-established, or published?

5. The cut-off of 4 interacting domains is arbitrary. More work is necessary to establish something that sounds 'reasonable'.

MINOR POINTS

1. Gene fusion events should cite the following: 10573422; the cited reference calls those Rosetta Stone proteins.

2. Low accuracy / low specificity: should be amended, since they implicitly refer to both a biological and a mathematical notion. The authors use the biological notion, therefore it should be absolutely clear that they do not refer to the mathematical notion.

3. Brun et al. reference in text should also include the following PMIDs: 12525261 / 14517352 / 14705023.

4. The authors need to clarify a little better their experimental design: although there is no overlap between the 13596 and 3812 human interaction pairs, it is not clear how do these correspond to the known human interactome. Is this some kind of jacknife test?

5. The sequence comparison part is very weak. The "sequence homology based method" of evaluation is not explained; the inference of interologs across species is not explained. More detail to ensure reproducibility of results should be provided.

**********
N.B. These are the comments made by the referee when reviewing an earlier version of this paper. Prior to publication the manuscript has been revised in light of these comments and to address other editorial requirements.