Domain Search

Find iBGCs by the protein domains they contain

The Domain search matches iBGCs on their protein domains — the functional units (Pfam/InterPro) that make up biosynthetic enzymes. Use it when you can describe a cluster by its machinery: “polyketide synthase plus a halogenase”, “an adenylation domain but no condensation domain”, or a specific ordered domain run.

Open the Domain Query builder from the domain chip. It has three modes, selected with the AND / OR / ARCH toggle.

The Domain Query builder. The AND / OR / ARCH toggle selects the mode; the box searches domains by accession or name.

Boolean modes: AND and OR

In AND/OR mode you build a set of domain conditions:

  1. Search for a domain by accession or name (for example PF00109, or KS). Matching domains list with their hit counts.
  2. Click one to add it as a condition. By default it is required.
  3. Click a condition to toggle it between required (+ req) and excluded (− excl), or remove it.

Then the mode decides how the required conditions combine:

  • AND — an iBGC must contain all required domains (and none of the excluded ones).
  • OR — an iBGC must contain any of the required domains (and none of the excluded ones).

This is the mode for presence/absence questions: “clusters with both a KS and an AT domain” (AND), or “clusters with any lanthipeptide cyclase” (OR). Excluded domains let you carve away families you are not interested in.

Architecture mode (ARCH)

Architecture mode matches the ordered arrangement of domains, not just their presence. Paste a comma-separated list of domain accessions in the order they occur along the protein/cluster, for example:

PF00109, PF02801, PF00501, PF08659

The search then ranks iBGCs by how closely their domain architecture resembles yours. Unknown accessions are silently dropped, and a token counter shows how many were parsed.

A weight slider balances the two ingredients of the architecture score:

  • towards Sørensen–Dice — reward sharing the same set of domains;
  • towards Adjacency Index — reward the same neighbouring pairs (the order/arrangement).

Set it towards Dice when membership matters most, towards Adjacency when the order of domains is the point. The default is balanced. See Scores & Metrics for how these combine.

Running and reading results

The search runs on Run Query and combines with any active filters and other conditions. Results sort by domain similarity, and the roster’s similarity column shows the domain match (Dice) score. On the Variables map you can put Domain match (Dice) on an axis.

Tips

  • Start broad (one or two AND domains) and tighten with excluded domains rather than adding many required ones at once.
  • Use the protein panel on a known cluster to read off the exact domain accessions, then paste them into architecture mode to find relatives.