Statistical bootstrapping for tandem mass spectral similarity
In phylogenetics, bootstrapping asks: if I slightly perturb my data, do I recover the same relationships? SpecReBoot asks the same question of MS/MS spectra — if I resample spectral features, do I recover the same matches and the same edges? Every result below carries a support value — match support for a library hit, edge support for a network edge — the fraction of replicates in which it held.
No spectrum matched yet
Paste a peak list or upload an MGF, then choose a library. Results come back in a few seconds.
What the columns mean
Each candidate is scored once on your spectrum as submitted, then re-scored on B resampled copies of it. The columns report both.
- Match support
- Fraction of replicates in which this candidate scored above the threshold. The reproducibility measure.
- Score
- Similarity on the original, unperturbed spectrum — what a conventional library search reports.
- Bootstrap mean ± sd
- Average score across replicates and its spread. A wide spread means the score depends on a few fragments.
- Top-1 / Top-5
- How often this candidate ranked first, or in the top five, across replicates.
A high score with low match support is the case worth catching: the match rests on evidence that resampling knocks out.
Re-ranking. Use the chips above the table, or click a column header, to rank by match support, score, or top-1/top-5 stability. The ▲▼ marker beside each rank shows how far that candidate moved against ranking by score alone — the places where the two disagree are the ones worth a second look.
Structures are drawn from the SMILES in the library record, where one is present. Click any structure to enlarge it.
Method described in Charria Girón et al. (2026), bioRxiv — please cite it if you use this module.
No network built yet
Upload an MGF to get the three SpecReBoot networks as GraphML, ready to open in Cytoscape.
What the three networks are
Edge support is the fraction of replicates in which two spectra came out as mutual top-k neighbours. The three networks are three readings of that one result.
- Base
- Similarity only — every pair above the similarity threshold, with no edge-support filter. This is the conventional network.
- Threshold
- Edges that clear both the similarity and the edge-support threshold. Fragile edges drop out here.
- Rescued
- The threshold network plus rescued edges: pairs whose similarity sits below the cut-off but which the bootstrap recovers again and again. These are the connections a similarity threshold alone would miss.
Nodes carry the compound name and precursor m/z; edges carry bootstrap_support and weight, so you can style by support directly in Cytoscape.
Method described in Charria Girón et al. (2026), bioRxiv — please cite it if you use this module.
Loading…
Contact
Please open a GitHub issue for bugs and feature requests.
Citing SpecReBoot
Please cite the paper for the module you used, both are available as preprints:
Charria Girón, E., Torres Ortega, L. R., Mergola Greef, J., Marin Felix, Y., Caicedo Ortega, N. H., Surup, F., Medema, M. H., & van der Hooft, J. J. J. (2026). Bootstrap resampling of mass spectral pairs with SpecReBoot reveals hidden molecular relationships. bioRxiv. doi.org/10.64898/2026.02.03.703446
Charria Girón, E., van IJcken, J., Della Vedova, L., Torres Ortega, L. R., & van der Hooft, J. J. J. (2026). Quantifying per-match reliability in library matching for untargeted metabolomics workflows. bioRxiv. doi.org/10.64898/2026.07.30.741704