The so-called ‘promiscuous’ domains which occur frequently in many
otherwise unrelated proteins, such as ATP-binding cassettes, actin
binding domains, WD repeats and SH3 domains (Marcotte et al.,
1999) were removed, to reduce errors.
• All proteins (fused and heterodimeric) identified in our study were
searched against InterPro (which combines diverse information
about protein families and domains from multiple databases)
(Hunter et al., 2011) for the full annotation of the individual protein
domains. The InterPro accession number for each protein domain is
indicated in the text by the three-letter code IPR followed by six digits.
• The predicted reference fused protein was split into its components
proteins and then checked by reverse BLAST (Altschul et al., 1997)
to assess whether these two proteins returned the initial two query
proteins as their best BLAST hit.