MIBiG & Validated BGCs

The experimentally characterised reference set that anchors novelty

Consult this page when you see an iBGC flagged Validated, or when you want to understand what novelty is measured against.

What MIBiG is

MIBiG (Minimum Information about a Biosynthetic Gene cluster) is a curated, public database of biosynthetic gene clusters that have been experimentally characterised. For each entry, a researcher has identified the cluster in a genome, linked it to a specific natural product through experimental evidence (gene knockout, heterologous expression, chemical isolation), and deposited the annotation — compound structure, bioactivity, literature references — in MIBiG.

It is the most widely used reference database for BGCs and is maintained by the natural products community. The platform loads a recent MIBiG release as its set of validated clusters.

What “validated” means here

In the platform, validated is applied to iBGCs that come from MIBiG (or another experimentally characterised reference set). A validated iBGC is one whose product is known. Everything else in the catalogue is a candidate — computationally predicted, but not yet experimentally tied to a product.

The distinction is binary, and deliberately so: it provides a clean reference line for novelty. Validated iBGCs carry a Validated badge in the roster and detail panels, and curated compound names and structures in their “compound features”.

How validated BGCs anchor novelty

Every novelty score is a distance from the nearest validated cluster:

  1. The platform measures each candidate iBGC’s similarity (from shared protein domains and their arrangement) to every validated iBGC.
  2. It keeps the highest similarity — the single closest validated cluster.
  3. Novelty is 1 − that similarity.

A novelty near 0 means the cluster closely matches a known one; a novelty near 1 means nothing validated resembles it. A validated cluster is its own closest validated match, so its novelty is 0 by construction.

Because novelty is relative to the current validated set, scores shift as MIBiG grows: as the community characterises new clusters, some previously “novel” candidates become less novel. This is expected, and reflects genuine progress.

Validated members in gene cluster families

The validated/candidate distinction also drives the value of a gene cluster family:

Validated members in family Reading
Zero The whole family is uncharacterised — the highest-value target.
One or a few You roughly know what the family makes; most members and their variants are unexplored.
Many Well characterised; lower discovery potential, though individual members may have unique modifications.

Limitations to keep in mind

MIBiG is the best available reference, but it is biased, and novelty scores inherit those biases:

  • Organism bias — well-studied taxa (especially Streptomyces) are over-represented, so clusters from them can look less novel than they are.
  • Class bias — polyketides and NRPs are far better represented than RiPPs or terpenes, which can inflate novelty for under-represented classes.
  • Temporal lag — newly published clusters take time to appear in MIBiG, so recent chemistry may not yet be a reference point.

Treat novelty as a relative signal, not an absolute truth, and corroborate with domain novelty and family context.

Tips

  • Validated clusters are good sanity checks. Inspect a MIBiG iBGC: it should score novelty 0 and sit among its relatives on the map.
  • Follow accessions to the literature. Note a validated cluster’s identifiers and look them up at mibig.secondarymetabolites.org for structures, bioactivities, and references.
  • “Validated” ≠ “important”. Many validated clusters make compounds of no known bioactivity; many candidates may make critical molecules that aren’t characterised yet.