Skip to content
Get packs

DeepSpot-M Spatial Transcriptomics

Predicts gene expression maps from H&E histology tiles using the DeepSpot-M model

Source Download skill
Repository
K-Dense-AI/scientific-agent-skills
Stars
55,780
Licence
MIT
Files
3 files

What it does

Once installed, the agent can help run DeepSpot-M, a research model that estimates gene activity (virtual spatial transcriptomics) from ordinary H&E stained tissue images. Given 224x224 pixel tiles at about 20x magnification, it outputs predicted log1p-CPM values for chosen genes, and it can guide tiling of whole slides and check tile size and resolution. The skill also explains licensing, gated model access, and the rules for valid inputs.

Great for

  • Academic or nonprofit researchers studying tumour histology and gene expression links
  • Bioinformaticians building pipelines that map predicted gene expression onto tissue images
  • Developers integrating an optional, gated research model into a larger project

Skip it if

  • Clinical or diagnostic decisions, since the skill states the model is for research use only and not clinical use
  • Commercial projects, because the code is noncommercial (PolyForm) and the weights are CC-BY-NC-SA with access restricted to noncommercial academic or nonprofit users
  • Measuring real gene expression, since outputs are predictions and not assays; results can be wrong for genes or tissues outside the training context

Try asking

  • “Set up DeepSpot-M in an isolated environment and check that my H&E tiles are 224x224 at about 20x”
  • “Predict log1p-CPM values for EPCAM, CD3D and MKI67 on these tile images and keep the gene labels aligned”
  • “Write a script that tiles a whole slide with histolab and exports predictions as an AnnData H5AD file”
  • “Explain which embedding source (evo2, orthrus, prott5, scgpt, apertus) to pick for my analysis”
Step-by-step guide: how to use DeepSpot-M Spatial TranscriptomicsInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 3 original files, the MIT licence and an attribution note.

    deepspot-m.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip deepspot-m.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip deepspot-m.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Verify magnification from the slide's physical pixel size (MPP) rather than the level number, and resample explicitly if needed
  • Keep the gene list beside the output columns, and check the gene symbols against model.gene_names before running
  • Confirm your licence eligibility on the live model access form before requesting weights

What's inside

  • references/api.md
  • references/whole_slide.md
  • SKILL.md

Redistributed unmodified from K-Dense-AI/scientific-agent-skills under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.