GCF-anchored and target-oriented mining of metagenomic BGC space for bioactive product discovery

Microbial natural products are vital for drug discovery, yet pervasive genomic fragmentation and data volume hinder the translation of metagenomic biosynthetic gene clusters (BGCs) into therapeutic leads. We present metasynBGC, a target-oriented framework that employs a "function-first" logic using experimentally validated gene cluster families (GCFs) as evolutionary templates and functional beacons. This dual-path strategy enables (i) template-guided reconstruction of fragmented BGCs via conserved biosynthetic synteny, and (ii) high-resolution prioritization of complete BGCs based on therapeutic potential. Applying metasynBGC to large-scale public metagenome-assembled genome (MAG) data sets, we uncovered hidden biosynthetic potential, including BGC0001089_plus, a novel bacillaene-like variant containing an additional functional pksF gene. Simultaneously, bioactivity-prioritized mining of nonribosomal peptide synthetase (NRPS) BGCs, coupled with deep learning and chemical synthesis, yielded six novel compounds. These molecules exhibited diverse in vitro cytotoxic profiles across seven cancer cell lines, with [Formula: see text] values as low as 38.04 µM. Compounds D and E showed potent activity and pronounced cell-line selectivity. By transforming microbial "dark matter" into an actionable reservoir of therapeutic leads, metasynBGC provides a practical target-oriented strategy for linking metagenomic BGC mining with downstream experimental validation. The scripts used in this study are publicly available at https://github.com/Shirly-Yang/metasynBGC.IMPORTANCEThe vast chemical diversity hidden within metagenomic data remains largely inaccessible because biosynthetic gene clusters (BGCs) are often highly fragmented. Our study introduces metasynBGC, a target-oriented framework designed to bridge these gaps by using well-characterized BGCs within gene cluster families (GCFs) as templates to reconstruct incomplete biosynthetic pathways. Unlike traditional methods, metasynBGC integrates a dual-path discovery engine: it performs priority-driven mining to precisely locate naturally intact clusters while reconstructing fragmented ones using representative BGC members as guides. We demonstrate its efficacy by identifying and synthesizing novel compounds with notable cytotoxic activity, providing a validated "sequence-to-molecule" pipeline. This work offers a scalable framework for the systematic discovery of therapeutic candidates from uncultivated microbes and expands our capacity to interpret genomic "dark matter." By enabling the functional validation of complex molecules from fragmented data, metasynBGC facilitates the translation of metagenomic data sets into promising leads for therapeutic development.