Key Concepts

The vocabulary you need to read the discovery platform

This page defines the terms used throughout the platform. Read it once before using the discovery platform; the Glossary gives shorter definitions for quick reference.

BGC — biosynthetic gene cluster

A biosynthetic gene cluster is a stretch of DNA whose genes work together to produce a natural product — an antibiotic, siderophore, pigment, toxin, or other small molecule. A typical BGC contains the core biosynthetic enzymes plus tailoring enzymes, regulators, and transporters. The platform’s job is to help you find BGCs whose products are likely to be new.

BGC prediction (source prediction)

The platform runs three detectors:

  • antiSMASH — rule-based detection with broad, well-known class definitions.
  • GECCO — a machine-learning detector.
  • SanntiS — a neural detector built on protein-domain annotations.

Each tool’s call on a contig is recorded as a source prediction, tagged with the tool that made it. Detectors frequently overlap or disagree on exact boundaries, which is why the platform integrates them.

iBGC — integrated BGC

An integrated BGC (iBGC) consolidates overlapping source predictions on the same contig into one candidate. It is the central unit of the platform: the roster lists iBGCs, the maps plot iBGCs, scores are computed per iBGC, and shortlists are made of iBGCs.

Each iBGC records the source predictions it was built from, the tools that called them, its location on the contig, its size, its predicted class, and its scores. An iBGC may be flagged:

  • Validated — it comes from an experimentally characterised reference (e.g. MIBiG). Validated iBGCs are the yardstick novelty is measured against.
  • Partial — the cluster runs off the edge of its contig, so it may be incomplete. Partial iBGCs are placed onto the maps by similarity to complete ones rather than being clustered directly.
  • Submitted — it came from an assembly you uploaded with Load Asset, not from the catalogue. Submitted iBGCs are shown in amber.

Assembly

An assembly is the genome or metagenome a BGC was found in — an isolate genome, a metagenome-assembled genome (MAG), or an environmental assembly. In the discovery platform the iBGC is primary; the assembly appears as context: its accession, organism, source collection, biome, and taxonomy show up as columns, filters, and detail fields. Some assemblies are flagged type strain (a reference strain available from a culture collection), which is useful when you need a culturable source.

Protein domains

A protein domain is a recognised functional unit within a protein — for example a ketosynthase or an adenylation domain. The pipeline annotates every gene’s domains using InterPro/Pfam. Domains are the platform’s main currency for comparing clusters: two iBGCs are considered similar when they share many of the same domains arranged in similar ways. Domain searches and the novelty scores both work from this domain vocabulary.

BGC class

The BGC class is a normalised label for the type of biosynthetic machinery — for example Polyketide (PKS), Non-ribosomal peptide (NRPS), RiPP, Terpene, Saccharide, or a hybrid. It tells you broadly what kind of chemistry the cluster encodes. See BGC Classes.

GCF — gene cluster family

A gene cluster family (GCF) is a group of iBGCs with similar domain content — a “natural product family” whose members likely make related compounds. The platform clusters all iBGCs into a hierarchy of families and gives each a dotted path (for example 42.7.3). GCF membership underlies domain novelty and is one of the strongest signals for prioritisation: a large family with no validated members is a widespread but uncharacterised biosynthetic strategy. See Gene Cluster Families.

Novelty and domain novelty

Two scores summarise how interesting an iBGC is:

  • Novelty — how far the iBGC is from the nearest validated cluster. 1 means nothing validated resembles it; 0 means it matches a known cluster exactly. Validated iBGCs score 0 by definition.
  • Domain novelty — the fraction of the iBGC’s protein domains that are unique within its own GCF. High values flag clusters carrying enzymes their relatives lack.

Both are shown as “—” when they cannot be computed (for example domain novelty for a single-member family). See Scores & Metrics.

Similarity

Wherever the platform compares two clusters — for novelty, for “find similar”, for a domain search — it measures similarity from shared protein domains and how those domains are arranged, not from raw DNA or protein alignment. This means clusters with the same biosynthetic logic can be matched even when their sequences have diverged. (The exception is Sequence Search, which deliberately aligns a query protein against cluster proteins.)

Predicted chemistry

The pipeline also predicts the chemistry of each cluster’s product. CHAMOIS assigns chemical classes from the ChemOnt ontology to the proteins in a cluster. Validated iBGCs additionally carry curated compound names and structures. Both appear in the iBGC detail panel as “compound features”.

Shortlist and report

As you work, you collect candidates in a shortlist (held in your browser). When ready you Generate Report: the platform materialises your selection into a standalone summary page with charts, tables, and downloads (GenBank, tables, JSON). See Shortlists and The Report.