Mass spectrometry imaging (MSI) excels at mapping metabolites and lipids with spatial precision, but the identity of detected ions often remains ambiguous. Investigators reported that single ions commonly match multiple candidate compounds. SpacePath was developed to keep those multiple assignments in the analysis pipeline rather than collapsing them to a single putative identity. The framework scores canonical pathways at each tissue location while preserving the full set of ion-to-compound matches. This permits pathway-level spatial scoring without discarding annotation uncertainty — a methodological choice that retains the real ambiguity inherent to MS1-based spatial metabolomics and lipidomics.
The study team observed that, when using standard mapping approaches, canonical pathway coverage of molecular assignments was very limited in MSI data. SpacePath’s combined representation — pathway scoring plus grouping of unmapped features (described below) — increased the share of assignments covered in pooled analyses. The authors reported quantitative coverage differences: pathway mappings alone covered on average only 1% of molecular assignments, whereas the SpacePath representation covered 24% of pooled assignments. The manuscript does not provide additional methodological parameters in the abstract; details of scoring algorithms and thresholds are available in the preprint and accompanying code repositories referenced by the authors.
Most detected metabolites and lipids do not map neatly into canonical biochemical pathways. To handle those unmapped assignments, the SpacePath team constructed de novo molecular families based on chemical similarity of the candidate compound lists linked to each ion. The pipeline groups MS1 features that lack pathway annotations into families, enabling spatial analysis of chemically related sets of ions despite absent pathway labels. Data indicated this approach brings structured interpretability to the sizable fraction of MSI signals that fall outside pathway databases, allowing investigators to interrogate spatial programs defined by related molecular species rather than database membership alone.
By keeping ion-to-compound ambiguity and grouping by chemical similarity, SpacePath preserves uncertainty while organizing data for downstream spatial analyses. The authors noted that this dual strategy — pathway scoring for mapped assignments and family creation for unmapped ones — extends the interpretive reach of MSI datasets without imposing false specificity on MS1 annotations.
The team benchmarked SpacePath on four public spatial lipidomics datasets. These included data from healthy mouse brain, cerebellum, a demyelination model, and human bladder cancer. Across these datasets, SpacePath recovered spatial patterns that had been reported previously in the literature. The study team observed concordance between SpacePath-derived spatial programs and known regional lipid distributions or disease-associated patterns in these reference datasets.
The preprint’s abstract does not enumerate exact performance metrics or per-dataset statistics beyond the recovered patterns and the pooled coverage figures noted above. For readers seeking implementation details and dataset-specific results, the authors provided supplementary material and links to data and code repositories referenced in the article footnotes.
Investigators applied SpacePath to newly acquired glioblastoma xenograft MSI datasets to explore tumor-associated lipid programs. The analysis identified lipid programs specific to the tumor mass and distinct programs in the surrounding brain tissue. The cohort showed spatial segregation of these programs when scored by SpacePath, suggesting the approach can resolve microenvironmental biochemical differences in tumor-bearing tissues.
To further test assignments emerging from de novo families, the authors used SpiderMass-MS/MS as an orthogonal validation method. SpiderMass-MS/MS independently confirmed four tumor-associated features that had been grouped into a de novo molecular family by SpacePath, and this MS/MS information refined the initial MS1-based assignments. The abstract does not list the chemical identities of those four features nor provide MS/MS spectra in the abstract itself; readers must consult the preprint and supplementary files for the spectral evidence and exact annotation refinements.
The authors deposited supporting resources: a Zenodo record and a GitHub repository were linked in the article footnotes for interested users to access code and data. The preprint appears on bioRxiv and has not been peer reviewed; the authors declared no competing interests. The manuscript documents an analytical approach intended to broaden spatial interpretation of MSI by preserving annotation uncertainty and assembling chemically coherent families for unmapped signals. Next steps and open questions described by the authors are available in the full preprint and supplementary materials, and implementation details for reproducing the analyses are provided in the linked code repositories.