Mass Spectra - Can They Reveal a Chemical Family

Archived study

Tests nine-class chemical-family prediction from processed positive-ion GNPS tandem mass spectra, with compound-separated training, validation and test sets. The selected population contained 2,000 compound groups and 3,849 spectra. Compound-holdout accuracy was 51.0% and macro-F1 0.477, compared with 0.107 for the precursor-only control.

Interpretation and limits

Identity separation does not establish scaffold, instrument or external-library generalization. The random-spectrum diagnostic uses a different population with identity overlap, so its higher score is not a causal estimate of leakage. Fitting-set ambiguities remain disclosed; source labels, fitting and bootstrap generation were not independently recreated.

Package and provenance

The unchanged research package comes from the upstream archive. It retains the original conversation and included scientific artifacts, including analysis code, results and figures. The poster is an AI-generated thematic illustration, not study evidence. This introduction summarizes archived findings; the analysis was not rerun. Later bounded checks do not certify the entire study or mean the original package was corrected.

Import provenance

This case was added to the Academic Research category on 2026-10-13, with the archived upstream case recorded as its source. The supplied session export and poster were uploaded without changing their contents. This introduction summarizes the archived findings and was standardized for the repository. The study, external data sources, and scientific results were not rerun or independently reverified for this import.

This GitHub catalog imports the case from the verified Hugging Face snapshot. The session export is referenced through that fixed revision; the cover and introduction are stored in this repository.

Case files

• Session export
• Case poster