← All apps
GNPS2-AnnEx
GNPS2 library annotation confidence dashboard. Enter a GNPS2 task ID (FBMN, Classical Networking, or Library Search) — matches are fetched over CORS and every confidence score is computed here in your browser. Your data never leaves your machine.
Scored scans
Click a row for the full metric breakdown. Download CSV
Scan
RT (min)
Precursor m/z
Top compound
Score
Label
Support
Total
Max cosine
Tanimoto
Entropy
← Back to scored scans
Scan —
RT (min) —
Precursor m/z —
Structure support —
Query entropy —
No SMILES or InChIKey on any hit in this scan — the score below is demoted (× , capped at ) because there is no structural evidence at all.
—
composite confidence score
—
—
—
Component
Weight
Normalised value
Raw
Contribution
Candidate structures in this scan
Biological context (MASST/REDU)
Consistent evidence: score ≥ 80
Inconclusive: 55–79.9
Inconsistent evidence: < 55
All computation runs client-side via RDKit.js (loaded from a CDN — cheminformatics of this depth has no pure-JS equivalent) and plain JavaScript. GNPS2 requests go directly from your browser to gnps2.org over CORS.
One exception: library rows whose only structure field is INCHI (no SMILES) are resolved via a best-effort, exact-match lookup to PubChem (structure identifier only, never your spectra/task data) before scoring, since no browser-side toolkit here can parse InChI — see the code comment near deriveSmilesFromInchi.
Biological context queries send only the InChIKey first block of each candidate structure to the MASST records API (masst-records.gnps2.org), via apps.gnps2.org's same-origin pass-through /api/masst_records — see MASST_RECORDS_URL.
Known difference from the Python app: on rare positional-isomer group ties (~1% of scans in testing), this port and the Python scorer can pick different, equally-valid tied groups — see the code comment near exactToPositional for details.