# Mass Spectrometry Imaging Data Loading

> Use when you have raw or preprocessed MS imaging data archived as an RDS file or from a Zenodo deposit that includes the full m/z feature set (e.g., 10,200 m/z values spanning 150–1000 m/z range) and spectrum count (e.

- Skill: `holobiomicslab/mass-spectrometry-imaging-data-loading` (Agent Skill)
- Install (CLI): `npx skillmds@latest add holobiomicslab/mass-spectrometry-imaging-data-loading`
- Raw SKILL.md: https://api.skillmd.com/api/skills/holobiomicslab/mass-spectrometry-imaging-data-loading/raw
- Safety review: PASS (external: skill-scanner PASS, skillspector PASS)
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- License: CC-BY-4.0
- Author: HolobiomicsLab (https://skillmd.com/u/holobiomicslab)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/holobiomicslab/mass-spectrometry-imaging-data-loading

---


# Mass Spectrometry Imaging Data Loading

## Summary

Load and initialize unprocessed MS imaging data (e.g., from Zenodo) into a Cardinal object, establishing the raw feature matrix, m/z axis, and spatial coordinates for downstream processing. This is the foundational step that materializes the high-dimensional m/z × spectrum matrix before normalization or peak refinement.

## When to use

You have raw or preprocessed MS imaging data archived as an RDS file or from a Zenodo deposit that includes the full m/z feature set (e.g., 10,200 m/z values spanning 150–1000 m/z range) and spectrum count (e.g., 4,959 spectra), and you need to load it into an R environment as a Cardinal object to begin the SpaMTP/Cardinal pipeline.

## When NOT to use

- Input is already a processed feature table (e.g., 687 cleaned peaks post-peakProcess) — use this skill only on unprocessed data.
- Data is in LC-MS format without spatial coordinates — this skill is specific to imaging data with (x, y) pixel information.
- Object is already instantiated in memory as a Cardinal or Seurat object — use this skill only for initial file I/O.

## Inputs

- Cardinal RDS file from Zenodo or local archive
- Unprocessed MS imaging data (imzML format or equivalent)
- URL or file path to raw imaging dataset

## Outputs

- Cardinal imaging object with m/z feature matrix
- Spatial coordinates (pixel grid) attached to object
- Raw m/z axis and spectrum intensities ready for downstream processing

## How to apply

Use Cardinal's data import functions (e.g., readRDS() with a Zenodo URL, or readImzML() for imzML files) to instantiate a Cardinal imaging object. Verify that the loaded object contains the expected number of m/z features and spectra (e.g., 10,200 m/z features, 4,959 spectra). The object should preserve the m/z axis (typically 150–1000 m/z range for metabolomics) and spatial coordinates (x, y pixel indices). This raw object is then passed downstream to summarizeFeatures(), TIC normalization, and peakProcess for peak refinement.

## Related tools

- **Cardinal** (Primary R package for MS imaging object instantiation and data structure; provides readRDS() and readImzML() import functions.) — https://github.com/Vitek-Lab/Cardinal3-vignettes
- **R** (Runtime environment for executing Cardinal loading commands and readRDS() file I/O.)
- **Zenodo** (Cloud repository hosting archived MS imaging datasets; Cardinal can load RDS files directly from Zenodo URLs.) — https://zenodo.org

## Examples

```
pig206 <- readRDS(url("https://zenodo.org/records/XXXXXX/files/pig206_unprocessed.RDS?download=1")); pig206
```

## Evaluation signals

- Verify object class is 'MSImageSet' or equivalent Cardinal imaging object.
- Confirm feature count matches expected m/z dimensionality (e.g., 10,200 features for pig206 dataset).
- Confirm spectrum count matches expected spatial replicates (e.g., 4,959 spectra for pig206).
- Inspect m/z axis range to confirm expected range (e.g., 150–1000 m/z for metabolomics).
- Verify spatial coordinates are present and non-degenerate (pixel grid intact).

## Limitations

- Loading may fail if Zenodo URL is expired or file format is corrupted; always verify URL accessibility before pipeline execution.
- Cardinal object structure assumes uniform m/z binning and rectangular pixel grid; non-standard imaging geometries may require custom preprocessing.
- Memory constraints apply for very large datasets (>50,000 spectra or >100,000 m/z features); consider subsampling or chunked loading for such cases.
- imzML files require auxiliary .ibd binary files to be co-located; loading will fail if .ibd is missing or inaccessible.

## Evidence

- [other] Load the unprocessed pig206 Cardinal object from Zenodo (10,200 m/z features, 4,959 spectra, m/z range 150–1000).: "Load the unprocessed pig206 Cardinal object from Zenodo (10,200 m/z features, 4,959 spectra, m/z range 150–1000)"
- [readme] Build on the foundation of a Seurat Class Object, this package has three major functionalities which include; (1) mass-to-charge ratio (m/z) metabolite annotation, (2) various downstream statistical analysis including differential metabolite expression and pathway analysis, and (3) integrative spatial-omics analysis.: "Build on the foundation of a Seurat Class Object, this package has three major functionalities which include; (1) mass-to-charge ratio (m/z) metabolite annotation, (2) various downstream statistical"