Skip to content

Representative domains & families

Overview

InterProScan often reports several overlapping member-database signatures for the same region of a protein, especially for domains. This is biologically useful because different databases may recognise the same feature in different ways, but the full result can be difficult to read at a glance. To make protein architectures easier to interpret, InterProScan automatically marks a representative subset of domain and family matches during a standard run.

JAG1_MOUSE domains

All domains found in JAG1_MOUSE

Representative matches provide a simplified view of the results, intended to highlight the main domain or family architecture of a protein without showing every overlapping signature. This is most useful for domains. For families, the value is often lower because family matches frequently describe most or all of the sequence rather than a local region.

JAG1_MOUSE representative domains

Representative domains in JAG1_MOUSE, giving a cleaner architecture summary

Info

Representative selection does not remove matches from the output. All matches and locations are still reported. In JSON, XML, and GFF3 output, representative locations are marked with representative=true, while other reported locations are marked with representative=false.

If you run InterProScan with --skip-repr-locations, all matches are still reported, but representative is set to false for every location.

How representatives are selected

Representative selection follows a simple principle: choose a small set of matches that summarizes the protein architecture while avoiding strong redundancy.

  1. InterProScan first identifies candidate matches for representative selection. These candidates are:

    • protein family matches from Pfam, CDD, PROSITE-profiles, SMART, NCBIFAM, CATH-Gene3D, and SUPERFAMILY
    • protein domain matches from Pfam, PANTHER, PIRSF, HAMAP, NCBIFAM, and SFLD

  2. Matches that overlap in the same region of the protein are grouped together.

  3. Within each overlapping group, InterProScan selects a combination of matches that covers as much of the sequence as possible while avoiding strongly redundant annotations.
  4. Matches that do not overlap with any alternatives are retained automatically.

In practice, this means the representative view is designed to retain the main biological signal of the architecture while suppressing repeated annotations of essentially the same region.