Guide · Agent skill
How to use the DepMap agent skill for cancer dependency analysis
Updated · written with AI from the project's own documentation and checked against our catalogue data
The DepMap skill gives an AI agent working knowledge of the Cancer Dependency Map, a public research database of gene knockout screens in cancer cell lines. Once installed, the agent can find and read DepMap release files, rank genes that cancer models depend on, compare biomarker groups, check whether genes behave together, and relate genetic dependencies to PRISM drug sensitivity data. This guide is for researchers, bioinformatics students and analysts who want to use it, and it explains how to install it, how to phrase requests, what it needs, and where its limits are.
DepMap Cancer Dependency AnalysisK-Dense-AI/scientific-agent-skills · MIT Download the skill Quick start
- 1Download and unzip the skill. Download the ZIP from the skill page and unzip it. Keep SKILL.md, scripts/depmap_data.py and references/dependency_analysis.md together in one folder.
- 2Place the folder where the agent can see it. Copy the folder into ~/.claude/skills/ for personal use, into a project's .claude/skills/ folder, or upload it under Settings, then Capabilities, then Skills in the Claude apps.
- 3Set up Python. Install Python 3.10 or newer with numpy, pandas and scipy 1.11 or newer so the bundled helpers can run.
- 4Ask a specific question. Describe the analysis, the cancer models or genes involved, and the DepMap release you want to use.
- 5Verify and record the files. Check the release metadata, obtain files through the portal where verification is required, compare MD5 checksums and save the release citation and README with your outputs.
What the skill makes the agent do
Without the skill, an agent has no specific guidance on how DepMap files are named, which identifiers they use, or which statistics they need. The skill adds that guidance. It covers CRISPR Chronos gene effects, cancer model annotations, omics biomarkers and PRISM drug sensitivity, and supports cancer-selective dependency, co-essentiality and candidate synthetic-lethality analyses.
The skill also asks the agent to record provenance. Saved outputs should include the release name, the retrieval date, the release citation, the local checksum and the release README. The skill states that its analyses generate target and synthetic-lethality hypotheses. They do not establish clinical efficacy, a therapeutic window or a causal interaction.
What you need before installing
The skill's bundled helper scripts need Python 3.10 or newer, with numpy, pandas and scipy 1.11 or newer. The workflow also needs network access to depmap.org for release discovery. Some downloads require manual verification in the DepMap portal in a browser, so the skill cannot complete every download on its own.
- Python 3.10+
- numpy, pandas and scipy 1.11 or newer
- Network access to depmap.org
- A browser for portal verification on some downloads
Install the DepMap skill from a ZIP file
Download the skill's ZIP file from the skill page and unzip it. For a personal install, place the unzipped skill folder in ~/.claude/skills/. For a single project, place it in that project's .claude/skills/ folder. In the Claude apps, you can instead upload the skill under Settings, then Capabilities, then Skills.
The skill folder should contain SKILL.md, scripts/depmap_data.py and references/dependency_analysis.md. Keep these files together so the agent can find the helper scripts and reference notes.
- Personal install: ~/.claude/skills/
- Project install: .claude/skills/ inside your project
- Claude apps: Settings, then Capabilities, then Skills, then upload
Trigger the skill with natural requests
You do not need a special command. Describe the analysis you want in plain language and name the cancer models, genes or drugs involved. The skill's scope covers ranking dependencies, comparing prespecified biomarker cohorts, inspecting co-essentiality and relating dependencies to PRISM compound response.
Good requests name the question, the data and the release. Asking for a stated release and its checksum makes the saved output easier to reproduce.
- Rank the genes that lung adenocarcinoma cell lines depend on most strongly in the current DepMap Public release, and record the release and checksum.
- Compare CRISPR gene effects for a target gene between models with and without a damaging mutation in a chosen gene.
- Check whether two candidate genes are co-essential across cancer models.
- Relate dependency on a gene to PRISM response to a compound across models.
Find a release and select its files
Start by discovering which releases are available. The skill's helper fetches the DepMap file catalogue, which lists each file's release, release date, filename and MD5 hash. Review the list before choosing a release, because the newest release is not guaranteed to have downloadable files yet.
After you obtain a file from the portal, the helper can compare its MD5 hash with the catalogue value. Note that some recent catalogue rows have blank download links, so availability must be checked per file.
from scripts.depmap_data import fetch_catalogue, select_file, verify_md5
catalogue = fetch_catalogue()
public = catalogue[catalogue['release'].str.startswith('DepMap Public ')]
print(public[['release', 'release_date']].drop_duplicates().sort_values('release_date', ascending=False).head(12))
release = 'DepMap Public 26Q1'
record = select_file(catalogue, release, 'CRISPRGeneEffect.csv')
print(record[['release', 'filename', 'md5_hash']])
# After obtaining this file from the portal:
# verify_md5('CRISPRGeneEffect.csv', record['md5_hash'])Read the data files correctly
Each DepMap file has its own structure, and joining files incorrectly is a common source of wrong results. The skill's reference lists the verified file names and what each one contains. For example, CRISPRGeneEffect.csv holds integrated, model-level Chronos effects with rows as ModelID and columns labelled as Symbol (EntrezID). The uncorrected effects file is not interchangeable with the corrected one.
Model.csv includes models that have no CRISPR measurements, so a missing value in the effect matrix is not a score of zero. For omics data, the skill asks for one row per ModelID, using the release's default-entry flag. Repeated conditions should not be averaged or counted as independent models without a clear design.
- Use the README of your chosen release to confirm units and identifier level.
- Do not concatenate the WGS and WES copy-number files as one cohort.
- Exclude metadata columns before calculating expression statistics.
- Treat damaging and hotspot mutation calls as answering different questions.
Interpret Chronos scores and dependency calls
Chronos gene effects are continuous and have no fixed bounds. More negative values mean stronger loss of fitness. In the normalised release matrix, about 0 is the nonessential-control anchor and about −1 is the common-essential-control anchor. A cutoff such as −0.5 is exploratory and is not a p-value, FDR or probability threshold.
Binary dependency calls depend on the statistic in the file you use. A high dependency probability and a low FDR are different rules, and the skill says not to convert silently between them. Positive effects may reflect growth advantage or technical noise, so they need validation.
What people use it for
- Ranking candidate gene dependencies across a chosen set of cancer cell lines for hypothesis generation.
- Comparing gene effects between prespecified biomarker groups, such as models with and without a damaging mutation.
- Checking whether two genes are co-essential, as a first step in a synthetic-lethality hypothesis.
- Relating a gene's dependency profile to PRISM drug sensitivity data for a research question.
- Teaching how DepMap files, identifiers and release metadata are structured.
Frequently asked questions
Is the DepMap agent skill free to use?
The skill's SKILL.md file lists a CC-BY-4.0 license, while the skill listing on this directory shows MIT. Check the license file in the downloaded folder before reuse. The license of the skill is separate from the terms of the DepMap data, which you should check for each release.
What do I need to run the DepMap skill?
You need Python 3.10 or newer, with numpy, pandas and scipy 1.11 or newer, plus network access to depmap.org. Some downloads need manual verification in a browser through the DepMap portal.
How do I install a Claude agent skill from a ZIP file?
Unzip the downloaded file and place the skill folder in ~/.claude/skills/ for personal use or in a project's .claude/skills/ folder. In the Claude apps, you can upload the skill under Settings, then Capabilities, then Skills.
Does the DepMap skill prove that a gene is a cancer drug target?
No. The skill says its results generate hypotheses for further research. They do not establish clinical efficacy, a therapeutic window or a causal interaction between genes.
Why does the DepMap skill ask me to check MD5 checksums?
A checksum confirms that the file you have matches the file recorded in the release catalogue. Saving the checksum with your results lets you show exactly which data produced them.
Are there alternatives to this skill for DepMap analysis?
The source does not compare this skill with other tools. You can work with the DepMap portal and its release files directly, but then the file handling, identifier checks and statistical checks are your responsibility.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if you are a cancer researcher, bioinformatics student or analyst who is comfortable with Python and statistics and wants release-pinned, reproducible DepMap analyses. Pick something else if you need clinical or treatment decisions, proof of causal gene interactions, or a simple data download, since the skill's own notes rule these out.
Download the skill