Translate GNPS2 FBMN annotations or Everything Bagel features into drug exposure readouts. Enter a public task ID and the app identifies its workflow from GNPS2 job metadata, or use your own files without a task ID. Analysis runs in your browser; task and Zenodo files are fetched directly by your browser.
Under development. This portable, single-file app can be opened locally to analyze uploaded FBMN tables offline. GNPS2 and Zenodo retrieval require internet access. Everything Bagel features, or your own feature table and query MGF, are searched against the Zenodo drug library in this browser.
InputsChoose a GNPS2 task or uploaded tables▾
Use a public GNPS2 job. Paste the task ID of a Feature-Based Molecular Networking (FBMN) or Everything Bagel job and click Run analysis. The app reads the job's metadata to pick its workflow. FBMN results are loaded directly; for Everything Bagel, the Library search step appears below. Needs internet access.
Use tables downloaded from an FBMN job — this works offline. Add featuretable_reformated.csv as the quantification table and merged_results_with_gnps.tsv as the annotation table, both from the same FBMN task, then click Run analysis. The quantification table needs row ID and .mzML/.mzXML sample columns; the annotation table needs #Scan#, SpectrumID, Compound_Name, MQScore and SharedPeaks.
Drop CSV/TSV here
or click to choose
Drop TSV/CSV here
or click to choose
Search your own spectra against the GNPS Drug Library — no task ID needed. Add a feature table with row ID and .mzML/.mzXML sample columns (for example an Everything Bagel aligned_features.csv) and a query MGF whose SCANS match those row IDs. Then use the Library search step below. Your files stay on your computer; only the drug reference spectra are downloaded.
Drop CSV/TSV here
needs row ID and .mzML/.mzXML sample columns
Drop MGF here
SCANS must match feature table row ID
Library search
Prepare the query spectra and the Zenodo drug reference spectra in this browser, then click Run analysis. The reference MGFs are large (hundreds of MB), so they download only when you select Prepare or Run, with progress shown above the results, and are then cached in this browser. The search runs here with the LibrarySearch_Rust engine (the one Everything Bagel uses), compiled to WebAssembly: an EB task's own search settings (precursor and fragment tolerance, minimum cosine and matched peaks, peak filtering) are read from its submission_parameters.yaml, a local MGF uses EB's defaults, and each feature keeps its best match. The first search indexes the drug library, which takes 10–20 seconds and about 2 GB of browser memory; later runs in the same tab reuse the index and take a few seconds.
Files
Analysis settings ⓘ
Drug classes come from the drug metadata built into this page. Fetch latest metadata loads the newest version from the GNPS Drug Library Zenodo record and caches it in this browser for offline use; this HTML file is not changed. Use embedded metadata switches back and clears that cache. The versions can differ and change readouts, so rerun analysis after switching.Drug metadata: the embedded offline copy
About the input files
GNPS2 task: public tasks are identified from submission_parameters.yaml before files are loaded. FBMN uses featuretable_reformated.csv and merged_results_with_gnps.tsv. Everything Bagel preparation retrieves aligned_features.csv and aligned_features_filled.mgf; MGF SCANS must match feature table row ID.
FBMN tables: the quantification table needs row ID and sample columns containing .mzML or .mzXML; the annotation table needs #Scan#, SpectrumID, Compound_Name, MQScore and SharedPeaks. Both should come from the same FBMN task.
Feature table + query MGF: a local library search needs no task ID. It takes a feature table with row ID and .mzML/.mzXML sample columns (for example an EB aligned_features.csv) and a query MGF whose SCANS match those row IDs.
🧭 Purpose
This app translates GNPS compound annotations into drug exposure information. It uses standardized pharmacologic vocabularies developed for all drug reference spectra on GNPS. By linking molecular features to pharmacological ontologies, it enables rapid identification of drugs present in your data and provides insight into their properties.
📘 How it works
For FBMN, enter a task ID with completed library annotations or upload its two result tables.
For library search, enter an Everything Bagel task ID, or skip the task ID and upload a feature table and query MGF. Select Prepare library search files: the browser obtains query spectra from GNPS2 or your files and drug reference spectra from Zenodo, showing download progress.
Once a search engine is installed, run library search and review the mapped drug readouts.
Review the visualizations and customize the annotation table; recalculate results after editing.
🧩 Key features
Annotates input feature tables with drug classifications.
Maps detected features to pharmacological ontologies across multiple levels.
Uses curated vocabularies to support biological and clinical interpretation.
🧪 Example dataset
Select Load example in the input panel to explore the app without preparing your own data.
This application is part of the GNPS downstream analysis ecosystem, MetaboApps. Uploaded files and results remain in your browser. Task and reference files are fetched from GNPS2 and Zenodo.
Edit the blue classification cells in the table, then apply edits to update every result. Clear the search first if you need to review all annotations.
Features excluded by default
Co-occurrence across samples: top bars count each exact class combination; left bars count samples containing each class; connected dots identify the classes in each combination.
Drug classes by sample
Therapeutic area → pharmacological class → parent compound. Flow width represents detections across samples.