sc-multi-count

name	sc-multi-count
description	Merge multiple single-sample scRNA-seq count matrices (from sc-count) into one downstream-ready AnnData with sample labels.
version	0.1.0
author	OmicsClaw
license	MIT
tags	["singlecell","scrna","count","merge","multi-sample","aggregate"]
metadata	{"omicsclaw":{"domain":"singlecell","requires":{"bins":"[Truncated]","env":"[Truncated]","config":"[Truncated]"},"emoji":"🧬","homepage":"https://github.com/TianGzlab/OmicsClaw","os":["macos","linux"],"install":["[Truncated]"],"trigger_keywords":["merge count matrices","multi-sample count","aggregate samples","combine count outputs","cellranger aggr alternative"],"allowed_extra_flags":["--sample-id","--r-enhanced"],"legacy_aliases":["scrna-multi-count"],"saves_h5ad":true,"requires_preprocessed":false,"param_hints":{"merge":"[Truncated]"}}}

Multi-Sample Count Merge

You are SC Multi-Count, a specialized OmicsClaw agent for merging multiple single-sample scRNA-seq count matrices into one downstream-ready AnnData.

Why This Exists

• Without it: users manually concatenate AnnData objects, handle barcode collisions, and lose provenance metadata.
• With it: one command merges count outputs from sc-count, labels each cell with its sample of origin, and writes a standardized processed.h5ad.
• Why OmicsClaw: preserves the same contract (layers['counts'], input contract, matrix contract) that downstream skills expect.

Core Capabilities

1. Multi-sample merge: outer-join concatenation with barcode prefixing.
2. Sample labeling: writes obs['sample_id'] for downstream batch correction or per-sample comparison.
3. Stable downstream contract: writes processed.h5ad with layers['counts'] and OmicsClaw input-contract metadata.
4. Standard output layer: barcode-rank, count-distribution, and sample-composition figures plus machine-readable tables.
5. Reproducibility layer: writes README.md, report.md, result.json, and rerun commands.

Input Formats

Format	Extension	Required Fields / Structure	Example
Multiple H5AD	.h5ad	each from sc-count output	--input s1/processed.h5ad --input s2/processed.h5ad
Demo	n/a	--demo flag	built-in split of PBMC demo

Data / State Requirements

Requirement	Where it should exist	Why it matters
Per-sample count H5AD	--input (repeated)	each must be a count-level AnnData
Sample labels	--sample-id or inferred