tooluniverse-single-cell

tooluniverse-single-cell

Production-ready single-cell genomics and expression matrix analysis skill for BixBench.

Overview

This skill provides comprehensive single-cell RNA-seq analysis using scanpy, anndata, scipy, and ToolUniverse. It handles both full scRNA-seq workflows (raw counts to annotated cell types) and targeted expression-level analyses (per-cell-type differential expression, gene correlation, clustering, ANOVA).

BixBench Coverage

18+ questions across 5 projects:

Project	Questions	Analysis Type
bix-22	5	Gene length vs expression correlation by immune cell type
bix-27	3	PCA on expression matrix, hierarchical/bootstrap clustering
bix-31	4	Cell-type-specific DE, t-tests comparing cell types
bix-33	2	Single-cell DE by cell type after treatment
bix-36	4	miRNA expression ANOVA across cell types, fold changes

Capabilities

Capability	Implementation
Data loading	h5ad, 10X, CSV/TSV count matrices via scanpy/anndata
Quality control	Cell/gene filtering, QC metrics, doublet detection
Normalization	Library-size normalization, log1p, z-score scaling
Dimensionality reduction	PCA (scanpy + sklearn), UMAP, t-SNE
Clustering	Leiden, Louvain, hierarchical, bootstrap consensus
Batch correction	Harmony, ComBat
Differential expression	Wilcoxon, t-test (scanpy), DESeq2 (pseudo-bulk)
Per-cell-type DE	Automated per-cell-type condition comparison
Statistical analysis	Pearson/Spearman correlation, t-test, Welch's, ANOVA
Cell-cell communication	OmniPath L-R interactions, CellPhoneDB, CellChatDB
Fold changes	Log2FC computation, median FC, distribution analysis
Multiple testing	BH, Bonferroni, BY correction
Cell type annotation	Marker gene scoring, ToolUniverse databases (HPA, MyGene)
Enrichment	gseapy ORA (GO, KEGG, Reactome, 220+ libraries)
Gene annotation	HPA, Ensembl, MyGene, UniProt via ToolUniverse

Required Packages

pip install scanpy anndata leidenalg umap-learn harmonypy gseapy pandas numpy scipy scikit-learn statsmodels

Optional:

pip install pydeseq2  # for pseudo-bulk DESeq2

Test Results

Total tests: 72
Passed: 72
Failed: 0
Success rate: 100.0%

Test Coverage

Test Group	Tests	Status
Data Loading	5	All pass
QC and Preprocessing	7	All pass
PCA	4	All pass
Clustering	4	All pass
Differential Expression	7	All pass
Correlation Analysis	5	All pass
Statistical Comparisons	9	All pass
Fold Change and DE Comparison	4	All pass
miRNA Analysis	4	All pass
Batch Correction	3	All pass
Cell Type Annotation	2	All pass
Enrichment Integration	2	All pass
ToolUniverse Integration	3	All pass
Edge Cases	7	All pass
Report Generation	2	All pass
Complete Pipeline	4	All pass

File Structure

tooluniverse-single-cell/
├── SKILL.md              - Main skill documentation (719 lines, redesigned 2026-02-17)
├── SKILL_OLD.md          - Original skill doc (2122 lines, backup)
├── QUICK_START.md        - Quick start guide with 8 examples
├── README.md             - This file
├── test_skill.py         - Test suite (72 tests, 100% pass rate)
├── REDESIGN_SUMMARY.md   - Redesign documentation
│
├── references/           - Detailed workflow guides (7 files)
│   ├── scanpy_workflow.md       - Complete scanpy pipeline (QC, normalize, cluster, DE)
│   ├── cell_communication.md    - OmniPath/CellPhoneDB integration (L-R interactions)
│   ├── clustering_guide.md      - Leiden, Louvain, hierarchical, bootstrap consensus
│   ├── marker_identification.md - Marker genes, cell type annotation
│   ├── trajectory_analysis.md   - Pseudotime, PAGA, trajectory inference
│   ├── seurat_workflow.md       - Seurat → Scanpy translation
│   └── troubleshooting.md       - Common errors and solutions
│
└── scripts/              - Utility scripts (3 files)
    ├── qc_metrics.py      - Calculate QC, apply filters
    ├── normalize_data.py  - Normalization pipeline
    └── find_markers.py    - Find markers, annotate cell types

Documentation Organization

• SKILL.md: High-level workflows, decision trees, BixBench patterns (start here)
• references/: Detailed implementation guides (dive deep when needed)
• scripts/: Ready-to-use utilities (for common operations)

Quick Start

New User? Start Here

1. Read SKILL.md - High-level overview and decision trees
2. Try examples in QUICK_START.md - 8 worked examples
3. Explore references/ - Detailed workflows when needed

Basic Workflow

import scanpy as sc

# Load data
adata = sc.read_h5ad("data.h5ad")

# Standard pipeline
sc.pp.normalize_total(adata, target_sum=1e4)
sc.pp.log1p(adata)
sc.pp.highly_variable_genes(adata, n_top_genes=2000)
sc.tl.pca(adata, n_comps=50)
sc.pp.neighbors(adata, n_pcs=30)
sc.tl.leiden(adata, resolution=0.5)

# Per-cell-type DE
for ct in adata.obs['cell_type'].unique():
    adata_ct = adata[adata.obs['cell_type'] == ct].copy()
    sc.tl.rank_genes_groups(adata_ct, groupby='condition',
                             groups=['treatment'], reference='control',
                             method='wilcoxon')
    df = sc.get.rank_genes_groups_df(adata_ct, group='treatment')
    n_sig = (df['pvals_adj'] < 0.05).sum()
    print(f"{ct}: {n_sig} DEGs")

Related Skills

• tooluniverse-rnaseq-deseq2: Bulk RNA-seq DESeq2 analysis
• tooluniverse-gene-enrichment: Gene enrichment and pathway analysis
• tooluniverse-statistical-modeling: Statistical regression and survival analysis
• tooluniverse-variant-analysis: VCF processing and variant annotation