Gene Cluster Families (GCF)

How iBGCs are grouped into families, and why it matters

A gene cluster family (GCF) groups iBGCs that share similar biosynthetic machinery — a “natural product family” whose members likely make related compounds. GCFs are one of the most useful concepts for prioritising candidates, and they underpin the domain novelty score.

What a GCF is

Think of a GCF as the BGC equivalent of a protein family: just as proteins with related sequences group into families, iBGCs with related domain content group into gene cluster families. Members of a family are expected to produce chemically related products, even when their exact structures differ.

Each GCF has a dotted path identifier reflecting its place in a hierarchy — for example 42.7.3, where each level is a finer grouping and the full path is the leaf family. Selecting a higher-level path includes all the families nested beneath it.

How families are built

The platform clusters all iBGCs from their protein domains and domain arrangement — the same composite similarity described in Scores & Metrics — not from raw sequence. In outline:

  1. Each iBGC’s domain content and neighbouring-domain pairs are compared to every other’s, giving a composite-Dice similarity.
  2. A nearest-neighbour graph links each iBGC to its most similar peers.
  3. A hierarchical community-detection step partitions that graph into nested families, producing the dotted leaf paths.

Because similarity is domain-based, the clustering groups iBGCs with the same biosynthetic logic even when their DNA or protein sequences have diverged beyond what alignment would detect — clusters with different gene order or gene count can still land in the same family.

Partial iBGCs (truncated at a contig edge) are not clustered directly; they are assigned to the family of their nearest complete neighbours.

Key family properties

  • Member count — how many iBGCs are in the family. Large counts mean a widespread biosynthetic strategy; a count of one is a singleton.
  • Validated members — how many family members are experimentally characterised (MIBiG). This tells you whether the family’s chemistry is known.

Why GCFs matter for discovery

GCF membership turns individual scores into decisions:

Large family, zero validated members. The highest-value pattern: a widespread biosynthetic strategy that no one has characterised. Many organisms carry it, so it is likely doing something biologically important — yet its product is unknown. Find a cluster here and you may open up a whole family of new compounds.

Singleton family (one member). The strongest novelty signal: nothing else in the catalogue is similar enough to group with it. High reward, but also high risk — with no relatives, there is no related chemistry to guide characterisation.

Family with some validated members. You roughly know what the family makes. Its unvalidated members are targeted candidates for variants of that known chemistry — useful if you want, say, a new analogue of a known antibiotic.

Small family, validated, type strain available. Lowest discovery potential: chemistry known, few carriers, and a culturable reference already exists.

GCFs in the discovery platform

  • Filter: the GCF chip restricts the catalogue to a family and its descendants. See Filtering.
  • iBGC detail: the GCF chip shows the cluster’s family path; click it to filter the whole discovery platform to that family — a quick pivot from one cluster to its relatives.
  • UMAP: points are coloured by family, so a tight single-colour clump is one GCF. See UMAP.
  • Report: the GCF distribution sunburst shows how a shortlist spreads across families. See The Report.
  • Domain novelty: computed within the leaf GCF — the fraction of a cluster’s domains unique among its family members.

Tips

  • After finding a promising cluster, click its GCF chip (or use Find similar iBGCs) to see the whole family and judge how crowded or sparse it is.
  • Check whether a family’s members come from diverse biomes and lineages (broadly important) or a single niche (rare but real).
  • Don’t dismiss small families — a handful of members with no validated relatives can be a genuine, under-explored strategy.