Ready to supercharge your AI agents?

Install any skill in seconds. Free to start, no credit card required.

Bulk RNA-seq batch correction with ComBat · FreedomIntelligence | DecimalAI

Bulk Rna Seq Batch Correction With Combat

SkillScore68Maintained

bulk-rna-seq-batch-correction-with-combat

Use omicverse's pyComBat wrapper to remove batch effects from merged bulk RNA-seq or microarray cohorts, export corrected matrices, and benchmark pre/post correction visualisations.

scienceby FreedomIntelligence

Links this skill to your agents and keeps it current with the author’s updates.

Automatic

Research pack

1 · One-time — install the CLI (docs ↗)

pip install decimalai

2 · Get the skill — writes .claude/skills/bulk-rna-seq-batch-correction-with-combat/SKILL.md

decimalai skills pull bulk-rna-seq-batch-correction-with-combat --out .claude/skills/

No CLI? Get the same file — curl · raw file

curl -fsSL --create-dirs -o .claude/skills/bulk-rna-seq-batch-correction-with-combat/SKILL.md https://app.decimal.ai/s/bulk-rna-seq-batch-correction-with-combat/SKILL.md

Raw file: /s/bulk-rna-seq-batch-correction-with-combat/SKILL.md ↗ · machine-readable summary at /s/bulk-rna-seq-batch-correction-with-combat.json ↗

In productionGlobal

how this skill behaves on real traffic · last 30 days

No production data for this skill yet — usage numbers appear here once agents start running it.

Anonymized aggregate across every org using this skill — proof from real production use, not a demo.

Does this skill help? · Benchmark: with vs. withoutsign in to run this on your agents

22 test cases from the publisher’s eval.yaml · each run twice on gemini-3.6-flash — with the skill injected, then without · last run 57d ago

Yes — +68% lift-2% tokens0% turns68% of whole cases pass with skill (15/22)

Test case	Without → With	Effect	Δ tokens	Δ turns
case-09	✗→✓	▲ Improved	—	—
case-13	✗→✓	▲ Improved	—	—
case-17	✗→✓	▲ Improved	—	—
case-20	✗→✓	▲ Improved	—	—
case-21	✗→✓	▲ Improved	—	—

Showing 5 of 22 cases

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v4 — 7/30/2026

live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing

IntentSafe

LLM judge · does the skill do what it claims

ContentClean

LLM judge · policy & harmful-content review

Source

Imported

FreedomIntelligence / OpenClaw-Medical-Skills

skills/bulk-combat-correction/SKILL.md

View on GitHub ↗

Imported 8/1/2026Version v1

Skill contents

SKILL.md

Bulk RNA-seq batch correction with ComBat

Overview

Apply this skill when a user has multiple bulk expression matrices measured across different batches and needs to harmonise them before downstream analysis. It follows t_bulk_combat.ipynb, w hich demonstrates the pyComBat workflow on ovarian cancer microarray cohorts.

Instructions

Import core libraries

Load omicverse as ov, anndata, pandas as pd, and matplotlib.pyplot as plt.
Call ov.ov_plot_set() (aliased ov.plot_set() in some releases) to align figures with omicverse styling.

Load each batch separately

Read the prepared pickled matrices (or user-provided expression tables) with pd.read_pickle(...)/pd.read_csv(...).
Transpose to gene × sample before wrapping them in anndata.AnnData objects so adata.obs stores sample metadata.
Assign a batch column for every cohort (adata.obs['batch'] = '1', '2', ...). Encourage descriptive labels when availa

ble.

Concatenate on shared genes

Use anndata.concat([adata1, adata2, adata3], merge='same') to retain the intersection of genes across batches.
Confirm the combined adata reports balanced sample counts per batch; if not, prompt users to re-check inputs.

Run ComBat batch correction

Execute ov.bulk.batch_correction(adata, batch_key='batch').
Explain that corrected values are stored in adata.layers['batch_correction'] while the original counts remain in adata.X.

Export corrected and raw matrices

Obtain DataFrames via adata.to_df().T (raw) and adata.to_df(layer='batch_correction').T (corrected).
Encourage saving both tables (.to_csv(...)) plus the harmonised AnnData (adata.write_h5ad('adata_batch.h5ad', compressio

n='gzip')).

Benchmark the correction

For per-sample variance checks, draw before/after boxplots and recolour boxes using ov.utils.red_color, blue_color, gree

n_color palettes to match batches.

Copy raw counts to a named layer with adata.layers['raw'] = adata.X.copy() before PCA.
Run ov.pp.pca(adata, layer='raw', n_pcs=50) and ov.pp.pca(adata, layer='batch_correction', n_pcs=50).
Visualise embeddings with ov.utils.embedding(..., basis='raw|original|X_pca', color='batch', frameon='small') and repeat fo

r the corrected layer to verify mixing.

Troubleshooting tips

Mismatched gene identifiers cause dropped features—remind users to harmonise feature names (e.g., gene symbols) before conca

tenation.

pyComBat expects log-scale intensities or similarly distributed counts; recommend log-transforming strongly skewed matrices.
If batch_correction layer is missing, ensure the batch_key matches the column name in adata.obs.

Examples

"Combine three GEO ovarian cohorts, run ComBat, and export both the raw and corrected CSV matrices."
"Plot PCA embeddings before and after batch correction to confirm that batches 1–3 overlap."
"Save the harmonised AnnData file so I can reload it later for downstream DEG analysis."

References

Tutorial notebook: t_bulk_combat.ipynb
Example inputs: omicverse_guide/docs/Tutorials-bulk/data/combat/
Quick copy/paste commands: reference.md

Illustrative Want to see what bulk-rna-seq-batch-correction-with-combat would put in front of your model? Preview the rendered body and check its trigger phrases — no LLM call, no signup.

Latest benchmark · publisher runVerified

7/26/2026, 10:21 UTC · 22 cases

Lift

+68%

68% vs 0%

Agent turns

1.0 vs 1.0

Tokens

-2%

2,216.68 vs 2,262.73

fail → pass · 15fail → fail · 7

Per-case results (22)

Case	Status	Duration (ms)			Turns			Tokens			Tool calls
Case	Status	Without	With	Δ	Without	With	Δ	Without	With	Δ	Without	With	Δ
case-09 After running bulk cohort harmonization, I need to export gene-by-sample expression DataFrames for both uncorrected and corrected values to CSV. How do I extract these DataFrames from the AnnData object?	fail→pass	—	—	—	—	—	—	—	—	—	—	—	—
case-10 I want to save my harmonized bulk transcriptomic AnnData object to disk in h5ad format for long-term storage. What compression setting should be specified when writing the file?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-13 I ran batch adjustment on my bulk transcriptomic dataset, but accessing adata.layers['batch_correction'] raises a KeyError. What is the most likely cause and fix for this error?	fail→pass	—	—	—	—	—	—	—	—	—	—	—	—
case-14 When merging two bulk RNA-seq matrices, the concatenated output drops almost all features. What step was likely missed during data preparation?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-17 Does running bulk batch correction in omicverse overwrite the original raw matrix in adata.X?	fail→pass	—	—	—	—	—	—	—	—	—	—	—	—
case-20 After concatenating individual bulk RNA-seq AnnData objects across batches, what verification step should be performed on the combined object prior to running correction?	fail→pass	—	—	—	—	—	—	—	—	—	—	—	—
case-21 How many principal components should be computed when calling omicverse PCA on raw and batch-corrected layers for bulk RNA-seq evaluation?	fail→pass	—	—	—	—	—	—	—	—	—	—	—	—
case-06 I just received raw paired-end FASTQ files from a bulk RNA-seq experiment. What pipeline should I use to align reads to the reference genome and quantify transcript abundances?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-07 I am loading bulk transcriptomic data from pickled Pandas DataFrames where rows are samples and columns are genes. How should I prepare these matrices before wrapping them into AnnData objects for bulk cohort harmonization?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-03 I need a Python script using omicverse to correct batch effects across two bulk microarray datasets. Please set up the AnnData objects with batch annotations, execute ComBat, store the corrected values in a dedicated layer, and produce boxplots comparing sample value distributions before and after harmonization.	fail→pass	—	—	—	—	—	—	—	—	—	—	—	—
case-18 When creating distribution boxplots for bulk RNA-seq cohorts across batches, I want to use omicverse custom batch color palettes. Which utility color objects should be passed?	fail→pass	—	—	—	—	—	—	—	—	—	—	—	—
case-01 I have expression matrices from three separate bulk RNA-seq cohorts stored as CSV files. I need to harmonize these cohorts using pyComBat in omicverse, merge them on intersecting features, run batch adjustment, and export the final count tables for both raw and corrected matrices along with the saved AnnData object.	fail→pass	—	—	—	—	—	—	—	—	—	—	—	—
case-02 I am analyzing multi-batch transcriptomic expression tables. Could you provide code to combine these cohorts, apply ComBat batch correction with omicverse, and generate before-and-after PCA embedding plots to evaluate batch mixing?	fail→pass	—	—	—	—	—	—	—	—	—	—	—	—
case-05 I am working with a single-cell RNA-seq dataset containing 50,000 cells across 4 biological replicates. How should I integrate these single-cell samples to remove batch effects?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-11 I want to compare PCA projections of bulk RNA-seq samples before and after batch adjustment using omicverse. How do I prevent PCA on uncorrected data from modifying or overwriting raw counts?	fail→pass	—	—	—	—	—	—	—	—	—	—	—	—
case-12 My raw bulk RNA-seq read count matrix has highly skewed non-negative integer distributions with large dynamic ranges. What preprocessing transformation should be applied before feeding the matrix to ComBat batch adjustment?	fail→pass	—	—	—	—	—	—	—	—	—	—	—	—
case-04 I have a gene expression count matrix from a bulk RNA-seq experiment with treatment and control samples. What is the standard workflow to identify differentially expressed genes between these conditions in Python?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-15 I am visualising PCA embeddings using omicverse functions to evaluate bulk sample clustering by batch. Which parameter should I set to display a small bounding frame around the embedding plot?	fail→pass	—	—	—	—	—	—	—	—	—	—	—	—
case-19 I am writing a script to remove batch effects from bulk microarray datasets. What specific omicverse function should be invoked for bulk batch adjustment?	fail→pass	—	—	—	—	—	—	—	—	—	—	—	—
case-16 When preparing separate bulk expression cohorts into individual AnnData objects before merging, how should batch metadata be recorded in each object?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-08 When combining three separate bulk RNA-seq matrices into a single AnnData object, standard concatenation might fill missing genes with NaNs or keep all outer genes. How do I ensure only shared genes present in all cohorts are retained?	fail→pass	—	—	—	—	—	—	—	—	—	—	—	—
case-22 Before generating diagnostic plots for bulk RNA-seq batch correction in omicverse, what setup function should be called at the beginning of the script to apply consistent styling?	fail→pass	—	—	—	—	—	—	—	—	—	—	—	—

These cases were also screened by the skill’s intent / safety pass.See the Trust review ↗Reproduce this benchmark in Colab ↗

How this skill was measured

DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +68 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.

Model measured: gemini-3.6-flash
Graded by: gemini-3.6-flash
Grading method: judged
Verification: verified
Pass rate without the skill: 0%
Pass rate with the skill: 68.2%
Cases attempted: 22
Cases counted toward lift: 22
Cases that regressed: 1
Skill version measured: v1
Run completed: 7/26/2026, 10:21 UTC

The per-case answers from this run were removed by the retention sweep, so the case table below shows the verdicts without the text either arm produced. The counts above were recorded at the time and are unaffected. Answers are now kept for 180 days.

SkillScore blends 4 signals1 of 4 complete

BenchmarkPass rate WITH skill on a verified run68

Live evalNeeds ≥10 evaluated traces in 30 days

AI ratingNeeds ≥10 rated traces

AdoptionNeeds ≥10 serves across ≥2 orgs in 30 days

Backed by 1 of 4 signals — each added signal makes the score more robust.

Verified on gemini-3.6-flash · re-tested 57 days ago

Other measured skills in the registry, with their headline benchmark lift.