Last reviewed: 2026-08-30
This roadmap describes public integration work. Current versions and evidence levels live in the tooling status.
The repository already documents or packages representative tools for:
- BGC calling: antiSMASH, plantiSMASH, DeepBGC;
- search: MMseqs2, cblaster, Foldseek, HMMER;
- comparison: clinker and JCVI MCScan;
- function evidence: InterProScan, DeepEC/ECPred, CLEAN/HIT-EC planning;
- review: igv-reports, Cytoscape.js, Quarto;
- reference data: MIBiG, Pfam, PMN access guidance.
Not every inventory entry has a runner. The packaged skill identifies the supported subset.
Add small public or synthetic fixtures for each wrapper. Record the command, version, output schema, expected failure behavior, and output hash.
Join BGC calls, cblaster results, clinker plots, and neighborhood tables through stable cluster identifiers. Preserve caller-specific evidence rather than collapsing disagreements.
Add a JCVI fixture with normalized feature identifiers, assembly-quality notes, anchors, plot output, and claim limits.
Define a compact schema for sequence similarity, domains, structure, EC prediction, confidence, and abstention. Keep each evidence channel visible.
Render one public campaign packet with the checked Quarto baseline and an exact Cytoscape.js version. Include static fallbacks and accessible figure descriptions.
Check documentation mirrors, local/private paths, credential patterns, generated runtime artifacts, large biological files, and wrapper syntax before publication.
A tool moves from planned or available to checked only when the repository contains:
- a public or synthetic fixture;
- a repeatable command;
- exact code, model, and database versions;
- a documented output contract;
- expected resource limits;
- license and redistribution notes;
- compact checked outputs or hashes.
Upstream popularity or an unpublished test is not enough for a public integration claim.