Comparative genome mining prioritizes low-reference-similarity biosynthetic gene-cluster families in a Streptomyces clavuligerus-enriched panel

Streptomyces clavuligerus is an industrial source of clavulanic acid and other specialized metabolites, yet much of its predicted biosynthetic capacity remains uncharacterized. We applied a uniform genome-mining workflow to a purpose-built panel of 100 publicly available Streptomyces assemblies: 19 S. clavuligerus assemblies representing 16 strain labels, 17 established metabolite producers, and 64 additional species. antiSMASH 8.0.4 predicted 3,841 biosynthetic gene-cluster (BGC) regions. At a prespecified operational cutoff of ≤30% similarity to the best KnownClusterBlast/MIBiG match, 2,139 regions (55.7%) were classified as low-reference-similarity candidates, including 488 regions with no reported match. Within this curated panel, assembly-blocked analysis agreed: 79 of 96 non-tied assemblies contained more than 50% candidates (one-sided exact sign-test P = 4.97 × 10⁻¹¹). The all-region fraction varied from 48.3% to 58.5% under ≤20% and ≤40% cutoffs, respectively, demonstrating threshold sensitivity. BiG-SCAPE 2.0.3 assigned the query regions to 2,070 gene-cluster families (GCFs), of which 1,960 lacked a MIBiG 3.1 reference member; 314 of these families were recurrent. All 19 S. clavuligerus assemblies contained KnownClusterBlast matches annotated as clavulanic acid/clavam, cephamycin C, holomycin, and tunicamycin clusters, and 51 MIBiG-unanchored GCFs were restricted to the focus group. BGC density remained higher than in the remainder of the curated panel after retaining one assembly per focus strain label. The resulting catalogue is a reproducible sequence-based prioritization resource, not evidence of chemical novelty, expression, or bioactivity.