Ready to supercharge your AI agents?

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

Bioinformatics Singlecell · FreedomIntelligence | DecimalAI

Bioinformatics Singlecell

SkillScore86Maintained

bioinformatics-singlecell

Advanced single-cell multi-omics analysis including scRNA-seq, scCITE-seq, scATAC-seq, and TARGET-seq. Use when analyzing single-cell data, cell type identification, trajectory analysis, differential expression, UMAP/clustering, integrating protein and RNA modalities (TotalVI), or working with Scanpy, Seurat, scvi-tools. Includes workflows for MPN, hematologic malignancies, megakaryocyte biology.

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/bioinformatics-singlecell/SKILL.md

decimalai skills pull bioinformatics-singlecell --out .claude/skills/

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

curl -fsSL --create-dirs -o .claude/skills/bioinformatics-singlecell/SKILL.md https://app.decimal.ai/s/bioinformatics-singlecell/SKILL.md

Raw file: /s/bioinformatics-singlecell/SKILL.md ↗ · machine-readable summary at /s/bioinformatics-singlecell.json ↗

In productionGlobal

how this skill behaves on real traffic · last 30 days

Runs

179

all versions · 30d

Eval pass

100%

global

Used when offered

100%

used / offered · all versions

By model· real traffic, every org · 30d — not the benchmark

Model	Eval pass	Runs
gemini-3.6-flash	100%	11

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

24 test cases from the publisher’s eval.yaml · each run twice on gemini-3.6-flash — with the skill injected, then without · last run 49d ago

Yes — +33% lift+23% tokens0% turns88% of whole cases pass with skill (21/24)

Test case	Without → With	Effect	Δ tokens	Δ turns
case-01	✗→✓	▲ Improved	23%	0%
case-02	✗→✓	▲ Improved	21%	0%
case-22	✗→✓	▲ Improved	-23%	0%
case-10	✗→✓	▲ Improved	30%	0%
case-11	✗→✓	▲ Improved	108%	0%

Showing 5 of 24 cases

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v4 — 8/1/2026

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/bioinformatics-singlecell/SKILL.md

View on GitHub ↗

License ProprietaryImported 8/1/2026Version v2Bundled 2 files

Skill contents

references/

SKILL.md

<!--

COPYRIGHT NOTICE

This file is part of the "Universal Biomedical Skills" project.

Copyright (c) 2026 MD BABU MIA, PhD <md.babu.mia@mssm.edu>

All Rights Reserved.

This code is proprietary and confidential.

Unauthorized copying of this file, via any medium is strictly prohibited.

Provenance: Authenticated by MD BABU MIA

-->

Single-Cell Multi-Omics Analysis

Core Libraries & Environment

python
# Essential imports
import scanpy as sc
import anndata as ad
import scvi
import muon as mu
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns

# Settings
sc.settings.verbosity = 3
sc.settings.set_figure_params(dpi=100, frameon=False, figsize=(6, 6))

Standard scRNA-seq Workflow

python
# 1. Load and QC
adata = sc.read_10x_mtx('path/to/filtered_feature_bc_matrix/')
sc.pp.filter_cells(adata, min_genes=200)
sc.pp.filter_genes(adata, min_cells=3)
adata.var['mt'] = adata.var_names.str.startswith('MT-')
sc.pp.calculate_qc_metrics(adata, qc_vars=['mt'], inplace=True)
adata = adata[adata.obs.pct_counts_mt < 20, :]

# 2. Normalization & HVG
sc.pp.normalize_total(adata, target_sum=1e4)
sc.pp.log1p(adata)
sc.pp.highly_variable_genes(adata, n_top_genes=2000, batch_key='batch')

# 3. Dimensionality reduction
sc.pp.scale(adata, max_value=10)
sc.tl.pca(adata, svd_solver='arpack')
sc.pp.neighbors(adata, n_neighbors=15, n_pcs=40)
sc.tl.umap(adata)
sc.tl.leiden(adata, resolution=0.5)

TotalVI for CITE-seq Integration

python
# Setup MuData
mdata = mu.MuData({'rna': adata_rna, 'protein': adata_prot})

# Train TotalVI
scvi.model.TOTALVI.setup_mudata(
    mdata, rna_layer='counts', protein_layer='counts',
    batch_key='batch', modalities={'rna_layer': 'rna', 'protein_layer': 'protein'}
)
model = scvi.model.TOTALVI(mdata, latent_distribution='normal', n_latent=20)
model.train(max_epochs=200, early_stopping=True)

# Get embeddings
mdata.obsm['X_totalVI'] = model.get_latent_representation()
sc.pp.neighbors(mdata, use_rep='X_totalVI')
sc.tl.umap(mdata)
sc.tl.leiden(mdata, key_added='leiden_totalVI', resolution=0.6)

Differential Expression

python
# DEG analysis
sc.tl.rank_genes_groups(adata, 'leiden', method='wilcoxon')
result = adata.uns['rank_genes_groups']
df = pd.DataFrame({
    'gene': result['names']['0'],
    'log2FC': result['logfoldchanges']['0'],
    'pval_adj': result['pvals_adj']['0']
})
sig_genes = df[(df['pval_adj'] < 0.05) & (abs(df['log2FC']) > 1)]

Publication-Quality Visualization

python
# Dot plot with proper expression cutoffs
sc.pl.dotplot(
    adata, var_names=marker_genes, groupby='leiden',
    expression_cutoff=0.0001, mean_only_expressed=False,
    standard_scale='None', smallest_dot=0.1, dot_max=1.0,
    cmap='viridis', colorbar_title='Expression'
)

# UMAP by batch
for batch in adata.obs['batch'].unique():
    adata_batch = adata[adata.obs['batch'] == batch]
    sc.pl.umap(adata_batch, color='FOXP3', title=f'{batch}')

Cell Type Annotation Markers

Hematopoietic Markers

HSC: CD34, KIT, THY1, CD38low
CMP/GMP: CD34+, CD38+, CD123
MEP: CD34+, CD38+, CD41/ITGA2B
Megakaryocytes: ITGA2B, PF4, GP1BA, PPBP, VWF
Erythroid: HBB, HBA1/2, GYPA, KLF1

MPN-Specific Markers

Inflammatory MKs: S100A8/9, CHI3L1, CXCL8, IL6
Fibrosis markers: TGFB1, COL1A1, LOXL2, VEGFA
Disease genes: JAK2, CALR, MPL, PPM1D, ASXL1

Output & Saving

python
# Save processed data
adata.write('processed_adata.h5ad')
model.save('totalvi_model/')
df.to_csv('DEG_results.csv', index=False)

See references/cell_markers.md for complete marker lists. See references/scvi_advanced.md for advanced scvi-tools workflows.

Illustrative Want to see what bioinformatics-singlecell 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

8/2/2026, 17:30 UTC · 24 cases

Lift

+33%

88% vs 54%

Agent turns

1.0 vs 1.0

Tokens

+23%

3,053.63 vs 2,483.08

fail → pass · 9pass → pass · 12fail → fail · 2pass → fail · 1

Per-case results (24)

Case	Status	Duration (ms)			Turns			Tokens			Tool calls
Case	Status	Without	With	Δ	Without	With	Δ	Without	With	Δ	Without	With	Δ
case-24 What filenames and paths should be used when writing out processed single-cell AnnData objects, saved totalVI integration models, and exported DEG tables?	pass→pass	12,800	2,722	-79%	1	1	0%	2,495	1,859	-25%	0	0	—
case-01 I have a 10x Genomics scRNA-seq dataset stored in a filtered feature barcode matrix folder, along with batch metadata. Could you write a Scanpy workflow in Python to execute quality control filtering, library size normalization, log transformation, batch-aware highly variable gene selection, PCA, neighbor graph construction, UMAP projection, and Leiden community detection? Please return the script that performs these steps and saves the processed AnnData object.	fail→pass	20,561	19,516	-5%	1	1	0%	4,584	5,649	+23%	0	0	—
case-02 We generated CITE-seq data with paired single-cell transcriptomics and surface antibody measurements across multiple experimental batches. Please provide a Python script using MuData and scvi-tools to setup totalVI, train the multi-modal integration model with early stopping, extract the latent representations into the multimodal object, calculate neighbors, build a joint UMAP, and cluster the cells.	fail→pass	14,941	12,499	-16%	1	1	0%	3,243	3,931	+21%	0	0	—
case-07 When performing initial quality control filtering on a raw single-cell AnnData object in Python, what cell gene count, gene cell count, and mitochondrial percentage filtering thresholds should be applied prior to normalization?	pass→fail	15,690	7,867	-50%	1	1	0%	3,227	2,877	-11%	0	0	—
case-21 List marker genes for identifying Megakaryocyte lineage cells and Erythroid lineage cells in single-cell RNA-seq studies.	pass→pass	14,170	12,790	-10%	1	1	0%	2,777	3,832	+38%	0	0	—
case-22 Which inflammatory marker genes specifically characterize inflammatory Megakaryocytes in myeloproliferative neoplasms (MPN)?	fail→pass	14,853	3,289	-78%	1	1	0%	2,570	1,968	-23%	0	0	—
case-23 What bone marrow fibrosis markers and disease driver genes are routinely evaluated in MPN multi-omics studies?	pass→pass	16,890	12,261	-27%	1	1	0%	3,333	3,777	+13%	0	0	—
case-03 I need to run differential gene expression analysis across Leiden clusters on my single-cell RNA-seq dataset. Please show me how to compute cluster marker genes using a Wilcoxon test with Scanpy, extract the gene names, fold changes, and adjusted p-values into a clean pandas DataFrame filtered for significantly dysregulated genes, and render a dot plot visualization of marker expression across clusters.	fail→fail	15,279	11,413	-25%	1	1	0%	3,089	4,007	+30%	0	0	—
case-04 I have a bulk RNA-sequencing count matrix from 12 tissue samples across two experimental conditions (control vs treatment). Please write an R script using DESeq2 to normalize counts, estimate dispersions, perform differential expression testing, and output significantly altered genes with adjusted p-value below 0.05.	pass→pass	12,838	14,922	+16%	1	1	0%	2,662	4,503	+69%	0	0	—
case-05 We performed single-cell V(D)J sequencing on T cells alongside 5' gene expression. How do I analyze TCR alpha and beta chain clonotype diversity, expansion, and CDR3 amino acid sequence alignment across clusters using Python?	pass→pass	16,678	20,227	+21%	1	1	0%	3,361	5,511	+64%	0	0	—
case-06 I have 10x Visium spatial transcriptomics slides from human kidney sections. How should I process the spatial matrix and high-resolution histology image to detect spatially variable genes using R or Python spatial frameworks?	fail→fail	16,641	18,056	+9%	1	1	0%	3,042	4,874	+60%	0	0	—
case-08 I am configuring highly variable gene selection for a multi-batch single-cell dataset. What target number of top genes and batch keyword parameter should be passed to the Scanpy HVG function?	pass→pass	8,695	5,210	-40%	1	1	0%	1,780	2,254	+27%	0	0	—
case-09 For feature scaling and principal component analysis on single-cell log-normalized expression data prior to neighbor graph construction, what maximum clipping value should be set during scaling, and which SVD solver engine is specified for PCA?	pass→pass	6,416	2,314	-64%	1	1	0%	1,197	1,788	+49%	0	0	—
case-10 What specific parameter values should be passed to Scanpy's neighbor finding (`n_neighbors` and `n_pcs`) and Leiden community detection (`resolution`) in standard scRNA-seq workflow processing?	fail→pass	12,135	10,310	-15%	1	1	0%	2,522	3,277	+30%	0	0	—
case-11 When calling the setup function for a totalVI model on a multimodal object containing transcriptomic and protein measurements across batches, what specific modal layer keys and batch key parameters must be supplied?	fail→pass	8,173	8,781	+7%	1	1	0%	1,501	3,128	+108%	0	0	—
case-12 What latent distribution specification and number of latent dimensions should be configured when instantiating the totalVI neural network model for CITE-seq integration?	pass→pass	9,842	4,858	-51%	1	1	0%	1,842	2,128	+16%	0	0	—
case-13 In totalVI multi-modal model training, what maximum epoch count and early stopping flag should be passed to the model training routine?	fail→pass	7,689	1,909	-75%	1	1	0%	1,534	1,680	+10%	0	0	—
case-14 After obtaining latent embeddings from a trained totalVI model, what key added name and cluster resolution value should be assigned during Leiden graph clustering?	fail→pass	9,751	1,878	-81%	1	1	0%	1,864	1,668	-11%	0	0	—
case-15 When parsing differential expression results from `adata.uns['rank_genes_groups']` into a pandas DataFrame for cluster '0', which specific dictionary sub-keys supply the gene names, log2 fold changes, and adjusted p-values?	pass→pass	4,900	3,384	-31%	1	1	0%	969	2,057	+112%	0	0	—
case-16 What exact adjusted p-value and log2 fold change thresholds should be used to filter differential expression results for significantly dysregulated genes?	pass→pass	10,642	6,029	-43%	1	1	0%	1,979	2,541	+28%	0	0	—
case-17 When rendering a marker gene dot plot using Scanpy, what specific values should be assigned to `expression_cutoff`, `smallest_dot`, and `dot_max`?	fail→pass	13,807	3,375	-76%	1	1	0%	2,411	2,019	-16%	0	0	—
case-18 What colormap, standard scaling, and colorbar title settings should be specified in Scanpy's marker dot plot function?	fail→pass	11,626	2,531	-78%	1	1	0%	2,088	1,859	-11%	0	0	—
case-19 How should Python code iterate over distinct batch identifiers in single-cell observation metadata to plot batch-specific UMAP projections of FOXP3 expression?	pass→pass	29,614	8,350	-72%	1	1	0%	2,846	2,943	+3%	0	0	—
case-20 Which surface protein markers characterize Hematopoietic Stem Cells (HSC) and Common Myeloid / Granulocyte-Monocyte Progenitors (CMP/GMP) in human bone marrow analysis?	pass→pass	13,194	9,249	-30%	1	1	0%	2,678	3,157	+18%	0	0	—

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. 24 cases were attempted. The headline lift of +33 percentage points is the difference between those two pass rates over the 24 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: 54.2%
Pass rate with the skill: 87.5%
Cases attempted: 24
Cases counted toward lift: 24
Cases that regressed: 1
Skill version measured: v2
Run completed: 8/2/2026, 17:30 UTC

Measured example — same prompt, with vs. without the skill · from the benchmark run · case case-01

Prompt

I have a 10x Genomics scRNA-seq dataset stored in a filtered feature barcode matrix folder, along with batch metadata. Could you write a Scanpy workflow in Python to execute quality control filtering, library size normalization, log transformation, batch-aware highly variable gene selection, PCA, neighbor graph construction, UMAP projection, and Leiden community detection? Please return the script that performs these steps and saves the processed AnnData object.

Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.

Show both answers

SkillScore blends 4 signals2 of 4 complete

BenchmarkPass rate WITH skill on a verified run88

Live evalNeeds ≥10 evaluated traces in 30 days

AI ratingNeeds ≥10 rated traces

AdoptionBreadth of real use across orgs84

Efficiency · separate axis (at equal correctness)+0% turns+30% tokens

Shown for transparency; not credited unless a verified run shows no correctness regression.

Backed by 2 of 4 signals — each added signal makes the score more robust.

Verified on gemini-3.6-flash · re-tested 49 days ago

Model	Method	Date	Lift
gemini-3.6-flash	verified	7/24/2026	+32%

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