spatial-preprocess

name	spatial-preprocess
description	Load matrix-level spatial transcriptomics data (including raw_counts.h5ad from spatial-raw-processing), run the current OmicsClaw scanpy-standard preprocessing workflow, and export a downstream-ready AnnData with explicit effective QC parameters plus a standardized preprocessing visualization contract.
version	0.5.0
author	OmicsClaw Team
license	MIT
tags	["spatial","preprocessing","QC","normalization","HVG","leiden","umap","visium","xenium"]
metadata	{"omicsclaw":{"domain":"spatial","allowed_extra_flags":["--data-type","--leiden-resolution","--max-genes","--max-mt-pct","--min-cells","--min-genes","--n-neighbors","--n-pcs","--n-top-hvg","--resolutions","--species","--tissue"],"param_hints":{"scanpy_standard":"[Truncated]"},"legacy_aliases":["preprocess"],"saves_h5ad":true,"requires_preprocessed":false,"requires":{"bins":"[Truncated]","env":"[Truncated]","config":"[Truncated]"},"emoji":"🔬","homepage":"https://github.com/TianGzlab/OmicsClaw","os":["macos","linux"],"install":["[Truncated]","[Truncated]"],"trigger_keywords":["preprocess","spatial preprocessing","spatial QC","normalize","visium","xenium","merfish","slide-seq","load spatial data","leiden","umap"]}}

🔬 Spatial Preprocess

You are Spatial Preprocess, the OmicsClaw skill for loading raw spatial transcriptomics data and converting it into a downstream-ready AnnData under the current scanpy_standard workflow. The wrapper now exposes a unified Python gallery, figure-ready exports, and optional R-side customization without changing the underlying preprocessing assumptions.

Why This Exists

• Without it: users manually combine loading, QC, normalization, HVG selection, PCA, neighbors, and clustering with inconsistent defaults.
• With it: one command produces a standardized processed.h5ad, QC-aware figures, clustering summaries, figure-ready CSV exports, and reproducibility helpers.
• Upstream boundary: if the user still has FASTQ pairs plus barcode-coordinate metadata, they should run spatial-raw-processing first.
• Why OmicsClaw: the wrapper preserves raw counts, records the actual effective QC thresholds, keeps platform-specific loading separate from the scientific preprocessing steps, and uses one stable output contract for downstream tools.

Core Capabilities

1. Current workflow: scanpy_standard: QC, normalization, seurat_v3 HVG selection, PCA, neighbors, UMAP, and Leiden clustering.
2. Multi-platform loading: Visium directories / h5 / h5ad, Xenium zarr / h5ad, and converted .h5ad inputs for MERFISH / Slide-seq / seqFISH.
3. QC filtering: spot or cell filtering using min_genes, min_cells, max_mt_pct, and optional max_genes.
4. Tissue-aware presets: optional wrapper-level presets for common tissues such as brain, heart, tumor, and lung.
5. Raw-count preservation: raw counts are stored in both adata.layers["counts"] and adata.raw before normalization.
6. Graph embedding: PCA, neighbor graph, and UMAP.
7. Clustering: primary Leiden clustering plus optional multi-resolution sweeps.
8. Standard Python gallery: emits a recipe-driven preprocessing gallery with clustering overview, QC diagnostics, PCA guidance, and threshold-context panels.
9. Figure-ready exports: writes figure_data/ CSVs plus a manifest so downstream tools can restyle the same preprocessing result without recomputing QC or clustering.