stereo-seq-quality-control-preprocessing

name: stereo-seq-quality-control-preprocessing description: Use when Stereo-seq or STOmics data needs QC, preprocessing, raw GEM/GEF/SAW loading, raw-to-count-matrix export, binning, bin/cell filtering, mitochondrial/count/gene QC maps, StereoPy output handling, GEM-to-Seurat conversion, or export of cleaned h5ad/RDS objects before downstream analysis.

Stereo-seq Quality Control Preprocessing

Use This For

• Loading Stereo-seq GEM/GEF/SAW/StereoPy outputs and creating binned or filtered analysis objects.
• Converting raw Stereo-seq GEM rows into a sparse gene-by-bin/cell count matrix plus coordinate metadata before downstream tools.
• Plotting QC metrics in tissue coordinates before domain, mapping, CCI, trajectory, or GRN analysis.
• Converting GEM-style count coordinates into Seurat-compatible spatial objects or tabular QC summaries.

Default Requirements

• Use bundled article-derived scripts in scripts/ before writing preprocessing code or searching external repositories.
• Read source_code.md and match templates by input object, spatial unit, and preprocessing goal. Do not hard-code tissue-to-tool rules.
• If no curated preprocessing entry fits, search code_candidates.tsv for additional article-linked repositories and reusable files before external search.
• Inspect local Python/R environments before running. Prefer conda run -n stereo-skills-py python ... for general Python scripts and conda run -n stereo-skills-r Rscript ... for R scripts. For StereoPy/GEF scripts that import stereo, use envs/environment-python-stereopy.yml if stereo is not already available. If stereo, Seurat, Matrix, or another required package is missing, stop and tell the user exactly what is missing and which preprocessing step is blocked.
• Preserve raw count outputs and record filtering thresholds.
• When converting raw GEM to a count matrix, export matrix, feature, barcode, coordinate, and provenance files; do not only create a plotting object.
• QC plots should use Arial, readable labels, equal-aspect tissue maps, and legends outside the data where possible.
• In the final response, state the reused paper, DOI, code repository or code DOI, original file, and dataset-specific edits.

Workflow

1. Identify the raw input type: GEF, GEM, h5ad, Seurat RDS, or SAW output folder.
2. Define the spatial unit and resolution: DNB, bin, pseudo-spot, or segmented cellbin.
3. For raw GEM-to-count-matrix work, run scripts/raw_gem_to_count_matrix_template.py first unless StereoPy/SAW output is required by the user.
4. Read source_code.md, then adapt the closest local script for GEF, Seurat, h5ad, or cellbin workflows.
5. Export cleaned object plus QC tables and figures.
6. Report thresholds, discarded units/genes if calculated, matrix dimensions, spatial-unit ID convention, and provenance.

Reusable Article Code

• scripts/endo_stereopy_qc_preprocessing_template.py: adapted from Endo.R StereoPy binning/QC for GEF-to-h5ad preprocessing.
• scripts/raw_gem_to_count_matrix_template.py: raw GEM-to-sparse-matrix/count-table exporter for gene-by-spatial-unit matrices, coordinates, QC metrics, and provenance.
• scripts/gf_gem_to_seurat_qc_template.R: adapted from GF/SPF cecum GEM-to-Seurat and spatial QC plotting.
• scripts/h5ad_spatial_qc_overview_template.py: adapted from Endo.R, human cortex, and GF/SPF cecum QC/stat plotting patterns for h5ad count/feature/spatial overview figures.

Output Expectations

• Cleaned h5ad/RDS or binned count object.
• Raw count matrix outputs when requested: Matrix Market .mtx, features.tsv, barcodes.tsv, coordinates.tsv, qc.tsv, and provenance JSON.
• QC tables for counts, detected genes, optional mitochondrial fraction, and coordinate bounds.
• Spatial QC maps and count/gene distributions.
• Filtering thresholds and whether they came from the template default or dataset-specific adaptation.
• Reused article code source, paper DOI, repository or code DOI, original file name, and dataset-specific edits.