spatial-deconvolution

A tool for estimating how much of each cell type is present in each spot of a spatial transcriptomics experiment, using a single-cell RNA sequencing reference. Spatial transcriptomics measures gene activity at locations in tissue, but one spot can contain several cells.

In plain words
What is it for?
Use it to map known cell types from single-cell data onto tissue locations and estimate their proportions across the tissue.
Why use it?
It helps recover cell-type information that is hidden when multiple cells share one measurement spot. It also provides several analysis methods for different data types and study needs.

Skill for Claude CodeCodex

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add skills/shangbiolab/spatialclaw/spatial-deconvolution
Any agent
npx skills add ShangBioLab/SpatialClaw --skill spatial-deconvolution
Clone the repo
git clone --depth 1 https://github.com/ShangBioLab/SpatialClaw

Made for: Claude Code, Codex.

Per session 28 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,309 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00028 $0.01309
Opus 5 $0.00014 $0.00655
Sonnet 5 $0.00006 $0.00262
Haiku 4.5 $0.00003 $0.00131

Measured 2d ago against content hash 1af56730b0cd, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

spatial-deconvolution scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

The scan reads SKILL.md. This mod also ships 2 executable files (spatial_deconvolution.py, tests/test_spatial_deconvolution.py), listed below but not scanned — reading those needs a real analyzer, not pattern matching.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

skills/spatial/spatial-deconvolution/SKILL.md · 164 lines

How it starts

The opening of the file, as written. The whole thing — 164 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Spatial Transcriptomics Deconvolution

Cell-type deconvolution for spatial transcriptomics: infer the proportion of each cell type per spot by mapping a single-cell RNA-seq reference onto the spatial data.

Supported methods: Tangram (default), Stereoscope, GraphST-guided.


Why This Exists

  • The Problem: Spatial platforms (Visium, Slide-seq…) capture multiple cells per spot, losing single-cell resolution.
  • The Solution: Align a high-resolution scRNA-seq reference to spatial data to estimate cell-type proportions per spot.
  • Why SPATIALCLAW: Wraps three complementary algorithms with a unified --method interface, handles gene-ID matching, and stores results under both method-specific and unified obsm keys.

Methods Overview

Method Algorithm Best For Key Dependency
tangram Deep-learning cell→space mapping (clusters mode) General purpose, GPU-friendly tangram-sc
stereoscope Negative-Binomial probabilistic model Count-data accuracy, multiple cell types scvi-tools
graphst GraphST spatial embedding + NNLS Noisy/sparse ST data, captures spatial context GraphST

Output Keys

All methods write to two obsm slots:

Key Content
<method>_ct_pred Method-specific proportions (e.g., tangram_ct_pred)
deconvolution_ct_pred Unified key — always present, same data

The unified deconvolution_ct_pred key lets downstream skills (visualisation, enrichment) work regardless of which method was used.


Workflow

  1. Load spatial (.h5ad) and scRNA reference (.h5ad).
  2. Intersect genes shared by both datasets.
  3. Deconvolve using the chosen method.
  4. Output updated spatial .h5ad with proportion matrices.

CLI Reference

# Tangram (default)
python spatialclaw.py run spatial-deconvolution \
  --input  <st_data.h5ad> \
  --reference <sc_data.h5ad> \
  --cell-type-key <column>

# Stereoscope
python spatialclaw.py run spatial-deconvolution \
  --input  <st_data.h5ad> \
  --reference <sc_data.h5ad> \
  --cell-type-key <column> \
  --method stereoscope \
  --n-epochs 10000

# GraphST-guided
python spatialclaw.py run spatial-deconvolution \
  --input  <st_data.h5ad> \
  --reference <sc_data.h5ad> \
  --cell-type-key <column> \
  --method graphst

Read the full file on GitHub · 164 lines

Files

What ships with it

2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 164 lines · 28 tokens per session scan A 1af56730b0cd

Subscribe to this mod's changes

spatial-deconvolution is a skill published in the GitHub repository ShangBioLab/SpatialClaw (11 stars, last pushed 3mo ago), licensed Apache-2.0. It adds 28 tokens to every session and 1,309 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Related

Other skills, from other repositories

Gene Panel Selection Workflow

End-to-end workflow for gene panel design in scRNA-seq and spatial transcriptomics, that should be STRICTLY followed: dataset understanding + smart downsampling + train/test splits, algorithmic selection (HVG/DE/RF/scGeneFit/SpaPROS), optimal sub-panel discovery (ARI vs size), biological completion with a stability…

aristoteleo/PantheonOS · 110 tokens

Developmental Gene Panel Design Workflow

Panel design for DEVELOPING / dynamic systems (embryonic organs, differentiation, regeneration). The target experiment is usually a LATE / terminal stage, but the biology is a trajectory: terminal cell types are end-products of earlier lineage programs. A panel built from the target stage alone resolves terminal…

aristoteleo/PantheonOS · 129 tokens

Spatial Omics Skills Index

Skills for spatial transcriptomics analysis including single-cell to spatial mapping (MOSCOT), 3D visualization (PyVista), and related spatial workflows.

aristoteleo/PantheonOS · 37 tokens

jts

Use when performing precise 2D computational geometry in Java — spatial predicates (contains, intersects), overlay operations, buffering, triangulation. JTS (Java Topology Suite): the canonical geometry engine used as blueprint for GEOS, Shapely, and NetTopologySuite.

znlgis/opengis-skills · 58 tokens

geometry-api-java

Use when working with Esri geometry types (Point, Polyline, Polygon, Envelope) in Java — spatial operations, JSON/GeoJSON/WKT I/O, geodesic calculations. Esri Geometry API for Java.

znlgis/opengis-skills · 49 tokens

bulkrna-trajblend

Load when placing bulk RNA-seq samples on a single-cell reference's pseudotime axis (NNLS deconvolution + nearest-neighbour mapping). Skip when plain cell-type proportions (use bulkrna-deconvolution); native single-cell trajectory inference (use sc-pseudotime).

TianGzlab/OmicsClaw · 65 tokens