sc-preprocessing

name	sc-preprocessing
description	Base scRNA preprocessing after QC: QC-aware filtering, normalization, highly variable gene selection, and PCA.
version	0.5.0
author	OmicsClaw
license	MIT
tags	["singlecell","scrna","preprocessing","qc","normalization","hvg","pca"]
metadata	{"omicsclaw":{"domain":"singlecell","allowed_extra_flags":["--method","--min-genes","--min-cells","--max-mt-pct","--n-top-hvg","--n-pcs","--normalization-target-sum","--scanpy-hvg-flavor","--pearson-hvg-flavor","--pearson-theta","--seurat-normalize-method","--seurat-scale-factor","--seurat-hvg-method","--sctransform-regress-mt","--no-sctransform-regress-mt","--r-enhanced"],"param_hints":{"scanpy":"[Truncated]","seurat":"[Truncated]","sctransform":"[Truncated]","pearson_residuals":"[Truncated]"},"legacy_aliases":["sc-preprocess"],"saves_h5ad":true,"requires_preprocessed":false,"requires":{"bins":"[Truncated]","env":"[Truncated]","config":"[Truncated]"},"emoji":"🧫","homepage":"https://github.com/OmicsClaw/OmicsClaw","os":["macos","linux"],"install":["[Truncated]"],"trigger_keywords":["single cell preprocess","scRNA preprocessing","normalize hvg pca","base preprocessing","Seurat preprocessing","SCTransform preprocessing"]}}

🧫 Single-Cell Base Preprocessing

This skill is the base preprocessing stage for scRNA data. It stops at a normalized, PCA-ready object so that users can choose the next branch explicitly:

• go to sc-batch-integration if multiple batches need correction
• or go directly to sc-clustering if batch correction is not needed

Why This Exists

• Without it: users manually chain QC-aware filtering, normalization, HVG selection, and PCA with inconsistent defaults.
• With it: one run produces a stable normalized, PCA-ready AnnData plus standard OmicsClaw outputs.
• Why OmicsClaw: the wrapper reuses shared single-cell canonicalization, QC, and filtering logic, and keeps a stable downstream contract.

Scope Boundary

Implemented methods:

1. scanpy
2. seurat
3. sctransform
4. pearson_residuals

Method-specific wrapper controls:

• scanpy: normalization_target_sum, scanpy_hvg_flavor
• seurat: seurat_normalize_method, seurat_scale_factor, seurat_hvg_method
• sctransform: sctransform_regress_mt
• pearson_residuals: pearson_hvg_flavor, pearson_theta

This skill does:

1. reuse existing QC state when available, otherwise compute the minimum needed QC metrics
2. filter cells and genes through the shared filtering logic
3. normalize or transform the expression matrix
4. select highly variable genes
5. compute PCA
6. export a normalized, PCA-ready processed.h5ad

This skill does not:

1. remove ambient RNA
2. remove doublets
3. perform batch integration
4. build the final neighbor graph, UMAP, or clusters
5. annotate cells or run DE

Input Expectations

• preferred input: raw-count-like AnnData or a QC-annotated count-oriented object
• if QC metrics already exist, they are reused instead of recomputed
• if the user has not reviewed QC yet and does not provide filtering thresholds, the recommended path is first