skills.rest
mims-harvard
mims-harvardOFFICIAL
Explore all skills by this author →
Website𝕏@marinkazitnikGitHub

tooluniverse-single-cell

Analyze single-cell RNA-seq data with scanpy from QC gating through clustering and annotation.

1.7k|254|Updated Mar 3, 2025
Search Tags:#single-cell#differential expression#clustering#Scanpy#scrna-seq#cell type annotation#Quality Control
One-click install
npx skills add https://github.com/mims-harvard/ToolUniverse --skill tooluniverse-single-cell
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: tooluniverse-single-cell
Source: https://github.com/mims-harvard/ToolUniverse/tree/main/plugins/tooluniverse/skills/tooluniverse-single-cell
Command: npx skills add https://github.com/mims-harvard/ToolUniverse --skill tooluniverse-single-cell

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires scanpy, anndata, pandas, numpy, scipy, scikit-learn, statsmodels, leidenalg, harmonypy, gseapy, and includes scripts (resource) and references (resource) components.

What problem does it solve?

Single-cell RNA-seq analysis requires many error-prone steps—QC gating, doublet detection, normalization, clustering, marker identification, and cell-type annotation—where wrong thresholds or skipped steps silently corrupt results. This Skill provides a complete, distribution-aware scanpy/AnnData workflow with explicit guidance on choosing thresholds, detecting doublets and ambient RNA, and interpreting results with statistical rigor.

Core Features & Use Cases

  • QC Gating & Filtering: Compute per-cell metrics (n_genes_by_counts, total_counts, pct_counts_mt, pct_counts_ribo), apply MAD-based outlier thresholds, and handle doublets (Scrublet/scDblFinder), ambient RNA (SoupX/DecontX), and empty droplets (EmptyDrops/knee).
  • Full Analysis Pipeline: Normalization, highly variable gene selection, PCA, UMAP/t-SNE, Leiden/Louvain clustering, marker gene identification, and cell-type annotation with Harmony batch correction.
  • Downstream Analysis: Per-cell-type differential expression (Wilcoxon), pseudo-bulk DESeq2, gene-property correlation, cell-cell communication via OmniPath ligand-receptor pairs, and trajectory analysis.
  • Use Case: Given an h5ad file of tumor-infiltrating immune cells, load it, gate out dying cells and doublets, cluster with Leiden, annotate T-cell and myeloid populations using CellMarker references, and identify which cell type has the most treatment-induced DEGs.

Quick Start

Ask the agent to load your h5ad or 10X scRNA-seq file, run QC gating with MAD-based thresholds, cluster the cells, and annotate cell types with marker genes.

Frequently Asked Questions about tooluniverse-single-cell

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run a complete scRNA-seq analysis pipeline with scanpy?▼

Load data with sc.read_h5ad or sc.read_10x_h5, calculate QC metrics and filter cells, normalize with sc.pp.normalize_total and log1p, select highly variable genes, run PCA, then cluster with sc.tl.leiden and visualize with UMAP. Finish with sc.tl.rank_genes_groups for marker identification.

How to choose QC thresholds for single-cell RNA-seq data?▼

Use MAD-based outlier detection instead of hardcoded cutoffs: flag cells beyond 5 median absolute deviations on log1p counts and genes, and 3 MADs upper-only on mitochondrial percentage paired with a biological ceiling. Always visualize violin and count-vs-mito scatter plots before committing to thresholds.

What is the difference between Leiden and Louvain clustering in scanpy?▼

Leiden is the recommended method because it guarantees connected communities and better optimization, while Louvain is faster but may produce disconnected clusters. Both accept a resolution parameter where higher values yield more clusters; start at 0.5 and adjust based on biology.

How do I detect doublets in scRNA-seq data?▼

Run Scrublet via sc.pp.scrublet on raw counts per sample before merging, with expected_doublet_rate around 0.8% per 1,000 cells recovered. Flag cells, cluster, then drop doublets, since real doublets form recognizable bridge clusters between parent cell types.

Should I use single-cell DE or pseudo-bulk DESeq2 for condition comparisons?▼

Pseudo-bulk DE (aggregating counts by sample, then running R DESeq2) is more statistically valid for condition comparisons because single-cell DE inflates significance by treating non-independent cells from the same sample as replicates. Use single-cell Wilcoxon tests mainly for marker gene finding.

Why do my marker genes appear in every cluster?▼

Ubiquitous marker expression usually indicates ambient RNA contamination, where cell-free mRNA from lysed cells is co-encapsulated into every droplet. Per-cell QC cannot detect this; apply SoupX or DecontX count correction using the raw and filtered matrices.

Related Skills

Explore More Education & Research Skills→
must1fmust1f

scanpy

Run end-to-end single-cell RNA-seq analysis with Scanpy on h5ad, 10X, or CSV datasets.

Community
Advanced
TianGzlabTianGzlab
155

spatial_enrichment

Perform ORA or GSEA pathway enrichment on spatial transcriptomics data.

Official
Intermediate
exon-researchexon-research
475

analytical-grounding

Retrieve pathway members, cell-type markers, and genomic interval overlaps from biological databases.

Official
Advanced
openaiopenai
5.3k

ngs-bulk-rnaseq

Routes bulk RNA-seq requests to count-generation QC or differential-expression workflows.

Official
Advanced