[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7EGTY6FSETZTRZXFHJXBEHQ":3,"skill-same:01M7EGTY6FSETZTRZXFHJXBEHQ":83},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":46,"commitSha":47,"skillDirectory":48,"forksCount":14,"repositoryHtmlUrl":49,"repositoryHomepageUrl":49,"ownerHtmlUrl":50,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":51,"install":52,"resources":64},"01M7EGTY6FSETZTRZXFHJXBEHQ","bio-workflows-scrnaseq-pipeline",null,"Orchestrates the end-to-end single-cell RNA-seq pipeline from 10x Cell Ranger output to annotated cell types, chaining ambient-RNA removal, doublet detection, MAD-adaptive QC, normalization, integration, clustering, marker annotation, and (separately) pseudobulk DE + differential abundance. Use when honoring the made-once counting commitments (reference build\u002FEnsembl vintage, --include-introns, cell-calling, feature namespace), ordering correct-then-detect-then-normalize (ambient before doublet before normalize), running per-sample QC before merge, integrating-then-clustering (never testing on integrated values), aggregating to pseudobulk for condition DE instead of cells-as-replicates, or pairing DE with differential abundance. Hands mechanism to the single-cell component skills; not a re-teach of any single step.","Mr-Nobody-Anonymous\u002FAll-skills","Mr-Nobody-Anonymous","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F184253119?v=4","01M7AEXFJHKTYQW5R0ZCQQTWFV","anonymous","01M7C00M3GQR4QKP9GZA42GFAK",0,"MIT","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fbiology\u002Fbio-workflows-scrnaseq-pipeline\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: bio-workflows-scrnaseq-pipeline\ndescription: Orchestrates the end-to-end single-cell RNA-seq pipeline from 10x Cell Ranger output to annotated cell types, chaining ambient-RNA removal, doublet detection, MAD-adaptive QC, normalization, integration, clustering, marker annotation, and (separately) pseudobulk DE + differential abundance. Use when honoring the made-once counting commitments (reference build\u002FEnsembl vintage, --include-introns, cell-calling, feature namespace), ordering correct-then-detect-then-normalize (ambient before doublet before normalize), running per-sample QC before merge, integrating-then-clustering (never testing on integrated values), aggregating to pseudobulk for condition DE instead of cells-as-replicates, or pairing DE with differential abundance. Hands mechanism to the single-cell component skills; not a re-teach of any single step.\ntool_type: mixed\nprimary_tool: Seurat\ngoal_approach_exempt: true\nworkflow: true\ndepends_on:\n  - single-cell\u002Fdata-io\n  - single-cell\u002Fpreprocessing\n  - single-cell\u002Fdoublet-detection\n  - single-cell\u002Fclustering\n  - single-cell\u002Fmarkers-annotation\nqc_checkpoints:\n  - after_loading: \"Expected cell count, reasonable UMI distribution\"\n  - after_qc: \"Remove low-quality cells and doublets\"\n  - after_normalization: \"No batch effects, HVGs look sensible\"\n  - after_clustering: \"Clusters are biologically meaningful\"\nsource: \"https:\u002F\u002Fgithub.com\u002FGPTomics\u002FbioSkills\"\nsource_repository: \"GPTomics\u002FbioSkills\"\nsource_path: \"workflows\u002Fscrnaseq-pipeline\u002FSKILL.md\"\nlicense: \"MIT\"\nimported_at: \"2026-09-20\"\n---\n\n## Version Compatibility\n\nReference examples tested with: Cell Ranger 10.0+ (human\u002Fmouse refs on Ensembl v110), Seurat 5.1+, Scanpy 1.10+, scDblFinder 1.16+, SoupX 1.6+, CellBender 0.3+, harmony\u002Fscvi-tools current, ggplot2 3.5+, numpy 1.26+\n\nBefore using code patterns, verify installed versions match. If versions differ:\n- Python: `pip show \u003Cpackage>` then `help(module.function)` to check signatures\n- R: `packageVersion('\u003Cpkg>')` then `?function_name` to verify parameters\n\nIf code throws ImportError, AttributeError, or TypeError, introspect the installed\npackage and adapt the example to match the actual API rather than retrying.\n\nNote: the Cell Ranger `filtered_feature_bc_matrix` is cell-CALLING only, NOT ambient-corrected; recovering low-RNA cells or running SoupX\u002FCellBender needs the RAW matrix. `--include-introns` is default TRUE since Cell Ranger 7 (essential for nuclei). `gex_only=True`\u002Fdefault silently drops Antibody Capture\u002FCRISPR\u002FHTO features. Confirm in-tool before quoting.\n\n# Single-Cell RNA-seq Pipeline\n\n**\"Analyze my single-cell RNA-seq data from counts to cell types\"** -> Orchestrate QC filtering, normalization (scanpy\u002FSeurat), batch integration (scVI\u002FHarmony), clustering, marker detection, cell type annotation, and trajectory inference.\n\nThis is a workflow skill: it owns the chaining decisions and hand-offs, not the internals of any one step. Every step below cross-references the component skill that teaches its mechanism.\n\n## Made-once commitments (decided at `cellranger count`, inherited downstream)\n\n| Commitment | Consequence inherited downstream |\n|------------|----------------------------------|\n| Reference build + Ensembl vintage (CR v10 = Ensembl v110) | Every gene ID\u002Fmarker lookup\u002Fcross-dataset merge; join on `gene_ids` (stable), not symbols (lossy across releases) |\n| `--include-introns` (default TRUE since CR 7) | Total UMI\u002Fcell and snRNA-vs-scRNA comparability; nuclei need introns; mixing intron-on\u002Foff samples is a hidden batch axis |\n| Cell-vs-empty-droplet calling | The filtered matrix is cell-CALLING only, NOT ambient-corrected; SoupX\u002FCellBender and low-RNA-cell rescue need the RAW matrix (filtered-only is irreversible loss) |\n| Feature\u002Fbarcode namespace (`gex_only`) | `gex_only=True` silently drops ADT\u002FHTO\u002FCRISPR features; set False and split by `feature_types` for CITE-seq\u002Fhashed pools |\n\n## Pipeline orchestration: the ordering decisions that make or break the result\n\nStage order is not arbitrary; getting it wrong silently corrupts every downstream step. The cross-cutting decisions this pipeline must get right:\n- Correct-then-detect-then-normalize, never reorder. Ambient-RNA removal (SoupX\u002FCellBender) reads the RAW droplet matrix and MUST precede doublet detection and the final QC thresholds (ambient inflation distorts both doublet scores and per-cell QC); doublet detection precedes integration (doublets seed fake intermediate clusters). See single-cell\u002Fpreprocessing.\n- Multiplexed (hashed) pools are demultiplexed FIRST. When samples are pooled with cell hashing (HTO\u002FMULTI-seq) or genotype multiplexing, assign each cell to its sample of origin and remove cross-sample doublets before any per-sample step; the per-sample QC and doublet operations below assume the pool is already split by sample. See single-cell\u002Fhashing-demultiplexing.\n- Per-sample QC and doublet detection come BEFORE any merge or integration. Ambient-RNA cleanup (SoupX\u002FCellBender), mito\u002Fgene-count filtering, and doublet calling (expected rate ~0.8% per 1,000 recovered cells) are per-capture operations; running them on a pooled object lets one lane's artifacts contaminate the shared null and lets surviving doublets form fake intermediate clusters. See single-cell\u002Fpreprocessing and single-cell\u002Fdoublet-detection.\n- Integrate, THEN cluster, THEN annotate - never reorder. For multi-sample designs, batch-correct on a shared embedding (Harmony\u002FscVI\u002FRPCA) and cluster on the corrected graph; clustering before correction makes clusters track samples or lanes instead of biology, and annotation has nothing to label until clusters exist. See single-cell\u002Fbatch-integration, single-cell\u002Fclustering, single-cell\u002Fcell-annotation.\n- Over-integration erases biology. Batch-mixing metrics (kBET, iLISI) are maximized by destroying structure, so a method that flattens real cell states scores perfectly; pair every batch metric with a bio-conservation metric and re-inspect rare populations after correction. Batch-corrected expression is for embedding and clustering only - never feed it to differential expression. See single-cell\u002Fbatch-integration.\n- Condition-level DE must use pseudobulk, not cells-as-replicates. Testing thousands of cells per donor as independent replicates is pseudoreplication and inflates false positives by orders of magnitude (Squair 2021); aggregate RAW counts per sample x cell type and hand them to DESeq2\u002FedgeR\u002Flimma-voom. Step 7 marker p-values are descriptive ranking for labeling, not condition inference. See single-cell\u002Fmarkers-annotation and differential-expression\u002Fdeseq2-basics.\n- Composition shifts are tested separately and masquerade as DE. A cluster-level expression change between conditions can be a pure proportion shift (a mixed cluster's substates re-balance with no gene changing per cell), invisible if only DE is run; always pair condition DE with a differential-abundance test (Milo cluster-free, or scCODA\u002Fsccomp\u002Fpropeller cluster-based). See single-cell\u002Fdifferential-abundance.\n\nThe Seurat and Scanpy paths below are written single-sample for clarity. A multiplexed design demultiplexes the pool first (single-cell\u002Fhashing-demultiplexing); a multi-sample design then inserts per-sample QC and doublet removal, a merge, and an integration step between normalization (Step 4) and dimensionality reduction (Step 5); the rest of the order is unchanged.\n\n## Workflow Overview\n\n```\n10X data (RAW + filtered_feature_bc_matrix)\n    |\n    v\n[0. Ambient removal] ---> SoupX\u002FCellBender on RAW (per sample)   (single-cell\u002Fpreprocessing)\n    |\n    v\n[1. Load Data] ---------> Read10X \u002F read_10x_h5\n    |\n    v\n[2. QC + Doublets] -----> MAD-adaptive nFeature\u002Fpercent.mt; scDblFinder\u002FScrublet (per sample, before merge)\n    |\n    v\n[3. Normalization] -----> SCTransform or LogNormalize\n    |\n    v\n[4. HVG Selection] -----> FindVariableFeatures\n    |\n    v\n[5. Dim Reduction] -----> PCA -> UMAP\n    |\n    v\n[6. Clustering] --------> FindNeighbors -> FindClusters\n    |\n    v\n[7. Markers] -----------> FindAllMarkers\n    |\n    v\n[8. Annotation] --------> Manual or automated\n    |\n    v\nAnnotated Seurat\u002FAnnData object\n```\n\n## Primary Path: Seurat (R)\n\nThe Seurat\u002FScanpy paths below start from the filtered matrix for clarity. In practice, run ambient-RNA removal FIRST on the RAW droplet matrix per sample (SoupX needs raw+filtered+clusters; CellBender folds calling+denoise and is best for snRNA\u002Fhigh-ambient) — pick ONE (double-correction over-strips). See single-cell\u002Fpreprocessing.\n\n### Step 1: Load 10X Data\n\n```r\nlibrary(Seurat)\nlibrary(ggplot2)\nlibrary(dplyr)\n\n# Load from Cell Ranger output\ndata_dir \u003C- 'cellranger_output\u002Ffiltered_feature_bc_matrix'\ncounts \u003C- Read10X(data.dir = data_dir)\n\n# Create Seurat object\nseurat_obj \u003C- CreateSeuratObject(counts = counts, project = 'my_project',\n                                  min.cells = 3, min.features = 200)\n```\n\n### Step 2: Quality Control\n\n```r\n# Calculate QC metrics\nseurat_obj[['percent.mt']] \u003C- PercentageFeatureSet(seurat_obj, pattern = '^MT-')\nseurat_obj[['percent.ribo']] \u003C- PercentageFeatureSet(seurat_obj, pattern = '^RP[SL]')\n\n# Visualize QC metrics\nVlnPlot(seurat_obj, features = c('nFeature_RNA', 'nCount_RNA', 'percent.mt'), ncol = 3)\n\n# Filter cells\nseurat_obj \u003C- subset(seurat_obj,\n                     nFeature_RNA > 200 &\n                     nFeature_RNA \u003C 5000 &\n                     percent.mt \u003C 20 &\n                     nCount_RNA > 500)\n\ncat('Cells after QC:', ncol(seurat_obj), '\\n')\n```\n\n**QC Checkpoint 1:** Review QC plots\n- Remove cells with very low\u002Fhigh gene counts\n- Remove cells with high mitochondrial content (dying cells)\n\n### Step 3: Doublet Detection\n\n```r\nlibrary(scDblFinder)\n\n# Convert to SCE for scDblFinder\nsce \u003C- as.SingleCellExperiment(seurat_obj)\nsce \u003C- scDblFinder(sce)\n\n# Add back to Seurat\nseurat_obj$doublet_class \u003C- sce$scDblFinder.class\nseurat_obj$doublet_score \u003C- sce$scDblFinder.score\n\n# Remove doublets\nseurat_obj \u003C- subset(seurat_obj, doublet_class == 'singlet')\ncat('Cells after doublet removal:', ncol(seurat_obj), '\\n')\n```\n\n### Step 4: Normalization with SCTransform\n\n```r\n# SCTransform (recommended for most analyses)\nseurat_obj \u003C- SCTransform(seurat_obj, verbose = FALSE)\n```\n\nAlternative: Standard normalization\n```r\nseurat_obj \u003C- NormalizeData(seurat_obj)\nseurat_obj \u003C- FindVariableFeatures(seurat_obj, selection.method = 'vst', nfeatures = 2000)\nseurat_obj \u003C- ScaleData(seurat_obj)\n```\n\nRegressing out `percent.mt` (or `nCount`\u002Fcell-cycle) is NOT reflexive: those covariates are confounded with real cell state and regressing them can erase biology, so only pass `vars.to.regress` for a covariate verified not confounded with the signal of interest. See single-cell\u002Fpreprocessing.\n\n### Step 5: Dimensionality Reduction\n\n```r\n# PCA\nseurat_obj \u003C- RunPCA(seurat_obj, npcs = 50, verbose = FALSE)\n\n# Determine optimal PCs\nElbowPlot(seurat_obj, ndims = 50)\n\n# UMAP\nn_pcs \u003C- 30  # Choose based on elbow plot\nseurat_obj \u003C- RunUMAP(seurat_obj, dims = 1:n_pcs, verbose = FALSE)\n```\n\n### Step 6: Clustering\n\n```r\n# Find neighbors\nseurat_obj \u003C- FindNeighbors(seurat_obj, dims = 1:n_pcs, verbose = FALSE)\n\n# Find clusters (try multiple resolutions)\nseurat_obj \u003C- FindClusters(seurat_obj, resolution = c(0.2, 0.4, 0.6, 0.8, 1.0), verbose = FALSE)\n\n# Visualize\nDimPlot(seurat_obj, reduction = 'umap', group.by = 'SCT_snn_res.0.4', label = TRUE)\n```\n\n**QC Checkpoint 2:** Assess clustering\n- Clusters should be visually separable on UMAP\n- Resolution 0.4-0.8 is often appropriate\n\n### Step 7: Find Marker Genes\n\n```r\n# Set identity to chosen resolution\nIdents(seurat_obj) \u003C- 'SCT_snn_res.0.4'\n\n# Find markers for all clusters\nmarkers \u003C- FindAllMarkers(seurat_obj, only.pos = TRUE, min.pct = 0.25, logfc.threshold = 0.25)\n\n# Top markers per cluster\ntop_markers \u003C- markers %>%\n    group_by(cluster) %>%\n    slice_max(n = 10, order_by = avg_log2FC)\n\n# Visualize top markers\nDoHeatmap(seurat_obj, features = top_markers$gene) + NoLegend()\n```\n\n### Step 8: Cell Type Annotation\n\n```r\n# Manual annotation based on known markers\n# Example for PBMC data:\ncluster_annotations \u003C- c(\n    '0' = 'CD4 T cells',\n    '1' = 'CD14 Monocytes',\n    '2' = 'B cells',\n    '3' = 'CD8 T cells',\n    '4' = 'NK cells',\n    '5' = 'CD16 Monocytes',\n    '6' = 'Dendritic cells'\n)\n\nseurat_obj$cell_type \u003C- cluster_annotations[as.character(Idents(seurat_obj))]\n\n# Final UMAP\nDimPlot(seurat_obj, reduction = 'umap', group.by = 'cell_type', label = TRUE)\n\n# Save object\nsaveRDS(seurat_obj, 'seurat_annotated.rds')\n```\n\n## Alternative Path: Scanpy (Python)\n\n```python\nimport scanpy as sc\nimport numpy as np\n\n# Load 10X data\nadata = sc.read_10x_h5('filtered_feature_bc_matrix.h5')\nadata.var_names_make_unique()\n\n# QC metrics\nadata.var['mt'] = adata.var_names.str.startswith('MT-')\nsc.pp.calculate_qc_metrics(adata, qc_vars=['mt'], percent_top=None, log1p=False, inplace=True)\n\n# Filter (flat cutoffs are illustrative; prefer MAD-adaptive, tissue-aware thresholds, see single-cell\u002Fpreprocessing)\nsc.pp.filter_cells(adata, min_genes=200)\nsc.pp.filter_genes(adata, min_cells=3)\nadata = adata[adata.obs.n_genes_by_counts \u003C 5000, :]\nadata = adata[adata.obs.pct_counts_mt \u003C 20, :]\n\n# Doublet detection (rate from recovered cells, not the 0.05 placeholder, see single-cell\u002Fdoublet-detection)\nexpected_rate = 0.008 * adata.n_obs \u002F 1000\nsc.pp.scrublet(adata, expected_doublet_rate=expected_rate)\nadata = adata[~adata.obs['predicted_doublet'], :]\n\n# Normalize and HVGs\nsc.pp.normalize_total(adata, target_sum=1e4)\nsc.pp.log1p(adata)\nsc.pp.highly_variable_genes(adata, n_top_genes=2000)\n\n# PCA, neighbors, UMAP -- stash log-normalized values in .raw BEFORE scaling, or rank_genes_groups\n# later computes logfoldchanges from z-scores (negative means -> NaN\u002Fgarbage LFCs)\nadata.raw = adata\nadata = adata[:, adata.var.highly_variable]\nsc.pp.scale(adata, max_value=10)\nsc.tl.pca(adata, n_comps=50)\nsc.pp.neighbors(adata, n_neighbors=15, n_pcs=30)\nsc.tl.umap(adata)\n\n# Clustering (pin the backend for reproducibility; default flips to igraph)\nsc.tl.leiden(adata, resolution=0.5, flavor='igraph', n_iterations=2, directed=False)\n\n# Markers\nsc.tl.rank_genes_groups(adata, 'leiden', method='wilcoxon')\nsc.pl.rank_genes_groups(adata, n_genes=10, sharey=False)\n\n# Save\nadata.write('scanpy_annotated.h5ad')\n```\n\n## Parameter Recommendations\n\n| Step | Parameter | Recommendation |\n|------|-----------|----------------|\n| QC | min.features | 200-500 |\n| QC | max.features | 2500-5000 (depends on data) |\n| QC | percent.mt | \u003C10-20% (tissue-dependent) |\n| SCTransform | vars.to.regress | none by default; only a validated non-confounded covariate |\n| PCA | npcs | 30-50 |\n| UMAP | dims | 15-30 (check elbow plot) |\n| Clustering | resolution | 0.4-0.8 (start with 0.5) |\n\nQC cutoffs above are illustrative starting points, not universal thresholds: set min\u002Fmax features and mito % per dataset from the data (MAD-adaptive, tissue-aware) rather than porting flat values, since healthy mito fraction varies by tissue. See single-cell\u002Fpreprocessing.\n\n## Troubleshooting\n\n| Issue | Likely Cause | Solution |\n|-------|--------------|----------|\n| All cells filtered | QC too strict | Relax thresholds |\n| Poor UMAP separation | Too few HVGs or PCs | Increase nfeatures, check n_pcs |\n| Too many\u002Ffew clusters | Wrong resolution | Adjust resolution parameter |\n| Unknown cell types | Missing markers | Check known marker genes manually |\n\n## Complete R Workflow\n\n```r\nlibrary(Seurat)\nlibrary(scDblFinder)\nlibrary(ggplot2)\nlibrary(dplyr)\n\n# Configuration\ndata_dir \u003C- 'filtered_feature_bc_matrix'\noutput_dir \u003C- 'results'\ndir.create(output_dir, showWarnings = FALSE)\n\n# Load\ncounts \u003C- Read10X(data.dir = data_dir)\nseurat_obj \u003C- CreateSeuratObject(counts = counts, min.cells = 3, min.features = 200)\ncat('Initial cells:', ncol(seurat_obj), '\\n')\n\n# QC\nseurat_obj[['percent.mt']] \u003C- PercentageFeatureSet(seurat_obj, pattern = '^MT-')\nseurat_obj \u003C- subset(seurat_obj, nFeature_RNA > 200 & nFeature_RNA \u003C 5000 & percent.mt \u003C 20)\ncat('After QC:', ncol(seurat_obj), '\\n')\n\n# Doublets\nsce \u003C- as.SingleCellExperiment(seurat_obj)\nsce \u003C- scDblFinder(sce)\nseurat_obj$doublet \u003C- sce$scDblFinder.class\nseurat_obj \u003C- subset(seurat_obj, doublet == 'singlet')\ncat('After doublet removal:', ncol(seurat_obj), '\\n')\n\n# Normalize (no reflexive vars.to.regress; regress only a validated non-confounded covariate)\nseurat_obj \u003C- SCTransform(seurat_obj, verbose = FALSE)\n\n# Dimension reduction\nseurat_obj \u003C- RunPCA(seurat_obj, npcs = 50, verbose = FALSE)\nseurat_obj \u003C- RunUMAP(seurat_obj, dims = 1:30, verbose = FALSE)\n\n# Cluster\nseurat_obj \u003C- FindNeighbors(seurat_obj, dims = 1:30, verbose = FALSE)\nseurat_obj \u003C- FindClusters(seurat_obj, resolution = 0.5, verbose = FALSE)\n\n# Markers\nmarkers \u003C- FindAllMarkers(seurat_obj, only.pos = TRUE, min.pct = 0.25, logfc.threshold = 0.25)\nwrite.csv(markers, file.path(output_dir, 'markers.csv'))\n\n# Save\nsaveRDS(seurat_obj, file.path(output_dir, 'seurat_object.rds'))\n\n# Plots\npdf(file.path(output_dir, 'umap.pdf'), width = 10, height = 8)\nDimPlot(seurat_obj, reduction = 'umap', label = TRUE)\ndev.off()\n\ncat('Pipeline complete. Object saved to:', output_dir, '\\n')\n```\n\n## Common Errors\n\n| Symptom | Cause | Fix |\n|---------|-------|-----|\n| Ambient markers everywhere (e.g. Hb across all PBMCs) | Ran SoupX\u002FCellBender on the FILTERED matrix (soup already removed) | Run ambient removal on the RAW droplet matrix, before QC |\n| Fake intermediate cell states | Doublets removed AFTER clustering\u002Fintegration | Call doublets per sample before merge\u002Fnormalize |\n| Clusters track sample\u002Flane, not biology | Clustered before integration | Integrate -> cluster -> annotate |\n| Inflated DE, thousands of false positives | Tested cells as replicates (on integrated values) | Pseudobulk RAW counts per sample x cell-type -> DESeq2\u002FedgeR\u002Flimma (Squair 2021) |\n| \"DE change\" is really a proportion shift | Only ran DE, not abundance | Pair with differential abundance (Milo\u002FscCODA\u002Fpropeller) |\n| Biology collapses to one blob | Reflexively regressed total_counts\u002Fcell-cycle | Regress only a validated, non-confounded covariate |\n| CITE-seq\u002Fhashtag features vanish | `gex_only=True` default | Set False; split by `feature_types` |\n\n## References\n\n- Heumos L, Schaar AC, Lance C, et al (2023) Best practices for single-cell analysis across modalities. *Nature Reviews Genetics* 24:550-572. DOI 10.1038\u002Fs41576-023-00586-w. (canonical pipeline order.)\n- Squair JW, Gautier M, Kathe C, et al (2021) Confronting false discoveries in single-cell differential expression. *Nature Communications* 12:5692. DOI 10.1038\u002Fs41467-021-25960-2. (cells-as-replicates is pseudoreplication; pseudobulk.)\n- Fleming SJ, Chaffin MD, Arduini A, et al (2023) Unsupervised removal of systematic background noise from droplet-based single-cell experiments using CellBender. *Nature Methods* 20:1323-1335. DOI 10.1038\u002Fs41592-023-01943-7. (ambient removal on the RAW matrix.)\n\n## Related Skills\n\n- database-access\u002Fgeo-data - Resolve GSE to SRA; detect SuperSeries before processing\n- database-access\u002Fsra-data - Download 10x records with --include-technical for barcodes\u002FUMIs\n- single-cell\u002Fdata-io - Loading 10X, h5ad, RDS, and h5mu formats\n- single-cell\u002Fpreprocessing - QC thresholds, ambient-RNA removal, normalization choice\n- single-cell\u002Fdoublet-detection - Per-sample doublet calling before integration\n- single-cell\u002Fhashing-demultiplexing - Assign multiplexed pools to samples and call cross-sample doublets before QC\n- single-cell\u002Fbatch-integration - Multi-sample integration and over-correction diagnosis\n- single-cell\u002Fclustering - Resolution sweep and cluster validation\n- single-cell\u002Fmarkers-annotation - Marker discovery, manual labeling, and pseudobulk condition DE\n- single-cell\u002Fcell-annotation - Automated reference-based label transfer\n- single-cell\u002Fdifferential-abundance - Test whether cell-type proportions shifted between conditions\n- single-cell\u002Ftrajectory-inference - Pseudotime and lineage reconstruction for continuous processes\n- single-cell\u002Fmultimodal-integration - CITE-seq and multiome joint analysis\n- differential-expression\u002Fdeseq2-basics - Pseudobulk condition DE engine for aggregated counts\n- differential-expression\u002Fde-results - Shrink, filter, and interpret pseudobulk DE results\n- pathway-analysis\u002Fgo-enrichment - Functional interpretation of marker and DE gene lists\n- workflows\u002Fgrn-pipeline - Downstream: infer gene regulatory networks (pySCENIC Path A) from the annotated object\n- workflows\u002Fspatial-pipeline - Downstream: the annotated reference deconvolves spatial spots\n","## Version Compatibility\n\nReference examples tested with: Cell Ranger 10.0+ (human\u002Fmouse refs on Ensembl v110), Seurat 5.1+, Scanpy 1.10+, scDblFinder 1.16+, SoupX 1.6+, CellBender 0.3+, harmony\u002Fscvi-tools current, ggplot2 3.5+, numpy 1.26+\n\nBefore using code patterns, verify installed versions match. If versions differ:\n- Python: `pip show \u003Cpackage>` then `help(module.function)` to check signatures\n- R: `packageVersion('\u003Cpkg>')` then `?function_name` to verify parameters\n\nIf code throws ImportError, AttributeError, or TypeError, introspect the installed\npackage and adapt the example to match the actual API rather than retrying.\n\nNote: the Cell Ranger `filtered_feature_bc_matrix` is cell-CALLING only, NOT ambient-corrected; recovering low-RNA cells or running SoupX\u002FCellBender needs the RAW matrix. `--include-introns` is default TRUE since Cell Ranger 7 (essential for nuclei). `gex_only=True`\u002Fdefault silently drops Antibody Capture\u002FCRISPR\u002FHTO features. Confirm in-tool before quoting.\n\n# Single-Cell RNA-seq Pipeline\n\n**\"Analyze my single-cell RNA-seq data from counts to cell types\"** -> Orchestrate QC filtering, normalization (scanpy\u002FSeurat), batch integration (scVI\u002FHarmony), clustering, marker detection, cell type annotation, and trajectory inference.\n\nThis is a workflow skill: it owns the chaining decisions and hand-offs, not the internals of any one step. Every step below cross-references the component skill that teaches its mechanism.\n\n## Made-once commitments (decided at `cellranger count`, inherited downstream)\n\n| Commitment | Consequence inherited downstream |\n|------------|----------------------------------|\n| Reference build + Ensembl vintage (CR v10 = Ensembl v110) | Every gene ID\u002Fmarker lookup\u002Fcross-dataset merge; join on `gene_ids` (stable), not symbols (lossy across releases) |\n| `--include-introns` (default TRUE since CR 7) | Total UMI\u002Fcell and snRNA-vs-scRNA comparability; nuclei need introns; mixing intron-on\u002Foff samples is a hidden batch axis |\n| Cell-vs-empty-droplet calling | The filtered matrix is cell-CALLING only, NOT ambient-corrected; SoupX\u002FCellBender and low-RNA-cell rescue need the RAW matrix (filtered-only is irreversible loss) |\n| Feature\u002Fbarcode namespace (`gex_only`) | `gex_only=True` silently drops ADT\u002FHTO\u002FCRISPR features; set False and split by `feature_types` for CITE-seq\u002Fhashed pools |\n\n## Pipeline orchestration: the ordering decisions that make or break the result\n\nStage order is not arbitrary; getting it wrong silently corrupts every downstream step. The cross-cutting decisions this pipeline must get right:\n- Correct-then-detect-then-normalize, never reorder. Ambient-RNA removal (SoupX\u002FCellBender) reads the RAW droplet matrix and MUST precede doublet detection and the final QC thresholds (ambient inflation distorts both doublet scores and per-cell QC); doublet detection precedes integration (doublets seed fake intermediate clusters). See single-cell\u002Fpreprocessing.\n- Multiplexed (hashed) pools are demultiplexed FIRST. When samples are pooled with cell hashing (HTO\u002FMULTI-seq) or genotype multiplexing, assign each cell to its sample of origin and remove cross-sample doublets before any per-sample step; the per-sample QC and doublet operations below assume the pool is already split by sample. See single-cell\u002Fhashing-demultiplexing.\n- Per-sample QC and doublet detection come BEFORE any merge or integration. Ambient-RNA cleanup (SoupX\u002FCellBender), mito\u002Fgene-count filtering, and doublet calling (expected rate ~0.8% per 1,000 recovered cells) are per-capture operations; running them on a pooled object lets one lane's artifacts contaminate the shared null and lets surviving doublets form fake intermediate clusters. See single-cell\u002Fpreprocessing and single-cell\u002Fdoublet-detection.\n- Integrate, THEN cluster, THEN annotate - never reorder. For multi-sample designs, batch-correct on a shared embedding (Harmony\u002FscVI\u002FRPCA) and cluster on the corrected graph; clustering before correction makes clusters track samples or lanes instead of biology, and annotation has nothing to label until clusters exist. See single-cell\u002Fbatch-integration, single-cell\u002Fclustering, single-cell\u002Fcell-annotation.\n- Over-integration erases biology. Batch-mixing metrics (kBET, iLISI) are maximized by destroying structure, so a method that flattens real cell states scores perfectly; pair every batch metric with a bio-conservation metric and re-inspect rare populations after correction. Batch-corrected expression is for embedding and clustering only - never feed it to differential expression. See single-cell\u002Fbatch-integration.\n- Condition-level DE must use pseudobulk, not cells-as-replicates. Testing thousands of cells per donor as independent replicates is pseudoreplication and inflates false positives by orders of magnitude (Squair 2021); aggregate RAW counts per sample x cell type and hand them to DESeq2\u002FedgeR\u002Flimma-voom. Step 7 marker p-values are descriptive ranking for labeling, not condition inference. See single-cell\u002Fmarkers-annotation and differential-expression\u002Fdeseq2-basics.\n- Composition shifts are tested separately and masquerade as DE. A cluster-level expression change between conditions can be a pure proportion shift (a mixed cluster's substates re-balance with no gene changing per cell), invisible if only DE is run; always pair condition DE with a differential-abundance test (Milo cluster-free, or scCODA\u002Fsccomp\u002Fpropeller cluster-based). See single-cell\u002Fdifferential-abundance.\n\nThe Seurat and Scanpy paths below are written single-sample for clarity. A multiplexed design demultiplexes the pool first (single-cell\u002Fhashing-demultiplexing); a multi-sample design then inserts per-sample QC and doublet removal, a merge, and an integration step between normalization (Step 4) and dimensionality reduction (Step 5); the rest of the order is unchanged.\n\n## Workflow Overview\n\n```\n10X data (RAW + filtered_feature_bc_matrix)\n    |\n    v\n[0. Ambient removal] ---> SoupX\u002FCellBender on RAW (per sample)   (single-cell\u002Fpreprocessing)\n    |\n    v\n[1. Load Data] ---------> Read10X \u002F read_10x_h5\n    |\n    v\n[2. QC + Doublets] -----> MAD-adaptive nFeature\u002Fpercent.mt; scDblFinder\u002FScrublet (per sample, before merge)\n    |\n    v\n[3. Normalization] -----> SCTransform or LogNormalize\n    |\n    v\n[4. HVG Selection] -----> FindVariableFeatures\n    |\n    v\n[5. Dim Reduction] -----> PCA -> UMAP\n    |\n    v\n[6. Clustering] --------> FindNeighbors -> FindClusters\n    |\n    v\n[7. Markers] -----------> FindAllMarkers\n    |\n    v\n[8. Annotation] --------> Manual or automated\n    |\n    v\nAnnotated Seurat\u002FAnnData object\n```\n\n## Primary Path: Seurat (R)\n\nThe Seurat\u002FScanpy paths below start from the filtered matrix for clarity. In practice, run ambient-RNA removal FIRST on the RAW droplet matrix per sample (SoupX needs raw+filtered+clusters; CellBender folds calling+denoise and is best for snRNA\u002Fhigh-ambient) — pick ONE (double-correction over-strips). See single-cell\u002Fpreprocessing.\n\n### Step 1: Load 10X Data\n\n```r\nlibrary(Seurat)\nlibrary(ggplot2)\nlibrary(dplyr)\n\n# Load from Cell Ranger output\ndata_dir \u003C- 'cellranger_output\u002Ffiltered_feature_bc_matrix'\ncounts \u003C- Read10X(data.dir = data_dir)\n\n# Create Seurat object\nseurat_obj \u003C- CreateSeuratObject(counts = counts, project = 'my_project',\n                                  min.cells = 3, min.features = 200)\n```\n\n### Step 2: Quality Control\n\n```r\n# Calculate QC metrics\nseurat_obj[['percent.mt']] \u003C- PercentageFeatureSet(seurat_obj, pattern = '^MT-')\nseurat_obj[['percent.ribo']] \u003C- PercentageFeatureSet(seurat_obj, pattern = '^RP[SL]')\n\n# Visualize QC metrics\nVlnPlot(seurat_obj, features = c('nFeature_RNA', 'nCount_RNA', 'percent.mt'), ncol = 3)\n\n# Filter cells\nseurat_obj \u003C- subset(seurat_obj,\n                     nFeature_RNA > 200 &\n                     nFeature_RNA \u003C 5000 &\n                     percent.mt \u003C 20 &\n                     nCount_RNA > 500)\n\ncat('Cells after QC:', ncol(seurat_obj), '\\n')\n```\n\n**QC Checkpoint 1:** Review QC plots\n- Remove cells with very low\u002Fhigh gene counts\n- Remove cells with high mitochondrial content (dying cells)\n\n### Step 3: Doublet Detection\n\n```r\nlibrary(scDblFinder)\n\n# Convert to SCE for scDblFinder\nsce \u003C- as.SingleCellExperiment(seurat_obj)\nsce \u003C- scDblFinder(sce)\n\n# Add back to Seurat\nseurat_obj$doublet_class \u003C- sce$scDblFinder.class\nseurat_obj$doublet_score \u003C- sce$scDblFinder.score\n\n# Remove doublets\nseurat_obj \u003C- subset(seurat_obj, doublet_class == 'singlet')\ncat('Cells after doublet removal:', ncol(seurat_obj), '\\n')\n```\n\n### Step 4: Normalization with SCTransform\n\n```r\n# SCTransform (recommended for most analyses)\nseurat_obj \u003C- SCTransform(seurat_obj, verbose = FALSE)\n```\n\nAlternative: Standard normalization\n```r\nseurat_obj \u003C- NormalizeData(seurat_obj)\nseurat_obj \u003C- FindVariableFeatures(seurat_obj, selection.method = 'vst', nfeatures = 2000)\nseurat_obj \u003C- ScaleData(seurat_obj)\n```\n\nRegressing out `percent.mt` (or `nCount`\u002Fcell-cycle) is NOT reflexive: those covariates are confounded with real cell state and regressing them can erase biology, so only pass `vars.to.regress` for a covariate verified not confounded with the signal of interest. See single-cell\u002Fpreprocessing.\n\n### Step 5: Dimensionality Reduction\n\n```r\n# PCA\nseurat_obj \u003C- RunPCA(seurat_obj, npcs = 50, verbose = FALSE)\n\n# Determine optimal PCs\nElbowPlot(seurat_obj, ndims = 50)\n\n# UMAP\nn_pcs \u003C- 30  # Choose based on elbow plot\nseurat_obj \u003C- RunUMAP(seurat_obj, dims = 1:n_pcs, verbose = FALSE)\n```\n\n### Step 6: Clustering\n\n```r\n# Find neighbors\nseurat_obj \u003C- FindNeighbors(seurat_obj, dims = 1:n_pcs, verbose = FALSE)\n\n# Find clusters (try multiple resolutions)\nseurat_obj \u003C- FindClusters(seurat_obj, resolution = c(0.2, 0.4, 0.6, 0.8, 1.0), verbose = FALSE)\n\n# Visualize\nDimPlot(seurat_obj, reduction = 'umap', group.by = 'SCT_snn_res.0.4', label = TRUE)\n```\n\n**QC Checkpoint 2:** Assess clustering\n- Clusters should be visually separable on UMAP\n- Resolution 0.4-0.8 is often appropriate\n\n### Step 7: Find Marker Genes\n\n```r\n# Set identity to chosen resolution\nIdents(seurat_obj) \u003C- 'SCT_snn_res.0.4'\n\n# Find markers for all clusters\nmarkers \u003C- FindAllMarkers(seurat_obj, only.pos = TRUE, min.pct = 0.25, logfc.threshold = 0.25)\n\n# Top markers per cluster\ntop_markers \u003C- markers %>%\n    group_by(cluster) %>%\n    slice_max(n = 10, order_by = avg_log2FC)\n\n# Visualize top markers\nDoHeatmap(seurat_obj, features = top_markers$gene) + NoLegend()\n```\n\n### Step 8: Cell Type Annotation\n\n```r\n# Manual annotation based on known markers\n# Example for PBMC data:\ncluster_annotations \u003C- c(\n    '0' = 'CD4 T cells',\n    '1' = 'CD14 Monocytes',\n    '2' = 'B cells',\n    '3' = 'CD8 T cells',\n    '4' = 'NK cells',\n    '5' = 'CD16 Monocytes',\n    '6' = 'Dendritic cells'\n)\n\nseurat_obj$cell_type \u003C- cluster_annotations[as.character(Idents(seurat_obj))]\n\n# Final UMAP\nDimPlot(seurat_obj, reduction = 'umap', group.by = 'cell_type', label = TRUE)\n\n# Save object\nsaveRDS(seurat_obj, 'seurat_annotated.rds')\n```\n\n## Alternative Path: Scanpy (Python)\n\n```python\nimport scanpy as sc\nimport numpy as np\n\n# Load 10X data\nadata = sc.read_10x_h5('filtered_feature_bc_matrix.h5')\nadata.var_names_make_unique()\n\n# QC metrics\nadata.var['mt'] = adata.var_names.str.startswith('MT-')\nsc.pp.calculate_qc_metrics(adata, qc_vars=['mt'], percent_top=None, log1p=False, inplace=True)\n\n# Filter (flat cutoffs are illustrative; prefer MAD-adaptive, tissue-aware thresholds, see single-cell\u002Fpreprocessing)\nsc.pp.filter_cells(adata, min_genes=200)\nsc.pp.filter_genes(adata, min_cells=3)\nadata = adata[adata.obs.n_genes_by_counts \u003C 5000, :]\nadata = adata[adata.obs.pct_counts_mt \u003C 20, :]\n\n# Doublet detection (rate from recovered cells, not the 0.05 placeholder, see single-cell\u002Fdoublet-detection)\nexpected_rate = 0.008 * adata.n_obs \u002F 1000\nsc.pp.scrublet(adata, expected_doublet_rate=expected_rate)\nadata = adata[~adata.obs['predicted_doublet'], :]\n\n# Normalize and HVGs\nsc.pp.normalize_total(adata, target_sum=1e4)\nsc.pp.log1p(adata)\nsc.pp.highly_variable_genes(adata, n_top_genes=2000)\n\n# PCA, neighbors, UMAP -- stash log-normalized values in .raw BEFORE scaling, or rank_genes_groups\n# later computes logfoldchanges from z-scores (negative means -> NaN\u002Fgarbage LFCs)\nadata.raw = adata\nadata = adata[:, adata.var.highly_variable]\nsc.pp.scale(adata, max_value=10)\nsc.tl.pca(adata, n_comps=50)\nsc.pp.neighbors(adata, n_neighbors=15, n_pcs=30)\nsc.tl.umap(adata)\n\n# Clustering (pin the backend for reproducibility; default flips to igraph)\nsc.tl.leiden(adata, resolution=0.5, flavor='igraph', n_iterations=2, directed=False)\n\n# Markers\nsc.tl.rank_genes_groups(adata, 'leiden', method='wilcoxon')\nsc.pl.rank_genes_groups(adata, n_genes=10, sharey=False)\n\n# Save\nadata.write('scanpy_annotated.h5ad')\n```\n\n## Parameter Recommendations\n\n| Step | Parameter | Recommendation |\n|------|-----------|----------------|\n| QC | min.features | 200-500 |\n| QC | max.features | 2500-5000 (depends on data) |\n| QC | percent.mt | \u003C10-20% (tissue-dependent) |\n| SCTransform | vars.to.regress | none by default; only a validated non-confounded covariate |\n| PCA | npcs | 30-50 |\n| UMAP | dims | 15-30 (check elbow plot) |\n| Clustering | resolution | 0.4-0.8 (start with 0.5) |\n\nQC cutoffs above are illustrative starting points, not universal thresholds: set min\u002Fmax features and mito % per dataset from the data (MAD-adaptive, tissue-aware) rather than porting flat values, since healthy mito fraction varies by tissue. See single-cell\u002Fpreprocessing.\n\n## Troubleshooting\n\n| Issue | Likely Cause | Solution |\n|-------|--------------|----------|\n| All cells filtered | QC too strict | Relax thresholds |\n| Poor UMAP separation | Too few HVGs or PCs | Increase nfeatures, check n_pcs |\n| Too many\u002Ffew clusters | Wrong resolution | Adjust resolution parameter |\n| Unknown cell types | Missing markers | Check known marker genes manually |\n\n## Complete R Workflow\n\n```r\nlibrary(Seurat)\nlibrary(scDblFinder)\nlibrary(ggplot2)\nlibrary(dplyr)\n\n# Configuration\ndata_dir \u003C- 'filtered_feature_bc_matrix'\noutput_dir \u003C- 'results'\ndir.create(output_dir, showWarnings = FALSE)\n\n# Load\ncounts \u003C- Read10X(data.dir = data_dir)\nseurat_obj \u003C- CreateSeuratObject(counts = counts, min.cells = 3, min.features = 200)\ncat('Initial cells:', ncol(seurat_obj), '\\n')\n\n# QC\nseurat_obj[['percent.mt']] \u003C- PercentageFeatureSet(seurat_obj, pattern = '^MT-')\nseurat_obj \u003C- subset(seurat_obj, nFeature_RNA > 200 & nFeature_RNA \u003C 5000 & percent.mt \u003C 20)\ncat('After QC:', ncol(seurat_obj), '\\n')\n\n# Doublets\nsce \u003C- as.SingleCellExperiment(seurat_obj)\nsce \u003C- scDblFinder(sce)\nseurat_obj$doublet \u003C- sce$scDblFinder.class\nseurat_obj \u003C- subset(seurat_obj, doublet == 'singlet')\ncat('After doublet removal:', ncol(seurat_obj), '\\n')\n\n# Normalize (no reflexive vars.to.regress; regress only a validated non-confounded covariate)\nseurat_obj \u003C- SCTransform(seurat_obj, verbose = FALSE)\n\n# Dimension reduction\nseurat_obj \u003C- RunPCA(seurat_obj, npcs = 50, verbose = FALSE)\nseurat_obj \u003C- RunUMAP(seurat_obj, dims = 1:30, verbose = FALSE)\n\n# Cluster\nseurat_obj \u003C- FindNeighbors(seurat_obj, dims = 1:30, verbose = FALSE)\nseurat_obj \u003C- FindClusters(seurat_obj, resolution = 0.5, verbose = FALSE)\n\n# Markers\nmarkers \u003C- FindAllMarkers(seurat_obj, only.pos = TRUE, min.pct = 0.25, logfc.threshold = 0.25)\nwrite.csv(markers, file.path(output_dir, 'markers.csv'))\n\n# Save\nsaveRDS(seurat_obj, file.path(output_dir, 'seurat_object.rds'))\n\n# Plots\npdf(file.path(output_dir, 'umap.pdf'), width = 10, height = 8)\nDimPlot(seurat_obj, reduction = 'umap', label = TRUE)\ndev.off()\n\ncat('Pipeline complete. Object saved to:', output_dir, '\\n')\n```\n\n## Common Errors\n\n| Symptom | Cause | Fix |\n|---------|-------|-----|\n| Ambient markers everywhere (e.g. Hb across all PBMCs) | Ran SoupX\u002FCellBender on the FILTERED matrix (soup already removed) | Run ambient removal on the RAW droplet matrix, before QC |\n| Fake intermediate cell states | Doublets removed AFTER clustering\u002Fintegration | Call doublets per sample before merge\u002Fnormalize |\n| Clusters track sample\u002Flane, not biology | Clustered before integration | Integrate -> cluster -> annotate |\n| Inflated DE, thousands of false positives | Tested cells as replicates (on integrated values) | Pseudobulk RAW counts per sample x cell-type -> DESeq2\u002FedgeR\u002Flimma (Squair 2021) |\n| \"DE change\" is really a proportion shift | Only ran DE, not abundance | Pair with differential abundance (Milo\u002FscCODA\u002Fpropeller) |\n| Biology collapses to one blob | Reflexively regressed total_counts\u002Fcell-cycle | Regress only a validated, non-confounded covariate |\n| CITE-seq\u002Fhashtag features vanish | `gex_only=True` default | Set False; split by `feature_types` |\n\n## References\n\n- Heumos L, Schaar AC, Lance C, et al (2023) Best practices for single-cell analysis across modalities. *Nature Reviews Genetics* 24:550-572. DOI 10.1038\u002Fs41576-023-00586-w. (canonical pipeline order.)\n- Squair JW, Gautier M, Kathe C, et al (2021) Confronting false discoveries in single-cell differential expression. *Nature Communications* 12:5692. DOI 10.1038\u002Fs41467-021-25960-2. (cells-as-replicates is pseudoreplication; pseudobulk.)\n- Fleming SJ, Chaffin MD, Arduini A, et al (2023) Unsupervised removal of systematic background noise from droplet-based single-cell experiments using CellBender. *Nature Methods* 20:1323-1335. DOI 10.1038\u002Fs41592-023-01943-7. (ambient removal on the RAW matrix.)\n\n## Related Skills\n\n- database-access\u002Fgeo-data - Resolve GSE to SRA; detect SuperSeries before processing\n- database-access\u002Fsra-data - Download 10x records with --include-technical for barcodes\u002FUMIs\n- single-cell\u002Fdata-io - Loading 10X, h5ad, RDS, and h5mu formats\n- single-cell\u002Fpreprocessing - QC thresholds, ambient-RNA removal, normalization choice\n- single-cell\u002Fdoublet-detection - Per-sample doublet calling before integration\n- single-cell\u002Fhashing-demultiplexing - Assign multiplexed pools to samples and call cross-sample doublets before QC\n- single-cell\u002Fbatch-integration - Multi-sample integration and over-correction diagnosis\n- single-cell\u002Fclustering - Resolution sweep and cluster validation\n- single-cell\u002Fmarkers-annotation - Marker discovery, manual labeling, and pseudobulk condition DE\n- single-cell\u002Fcell-annotation - Automated reference-based label transfer\n- single-cell\u002Fdifferential-abundance - Test whether cell-type proportions shifted between conditions\n- single-cell\u002Ftrajectory-inference - Pseudotime and lineage reconstruction for continuous processes\n- single-cell\u002Fmultimodal-integration - CITE-seq and multiome joint analysis\n- differential-expression\u002Fdeseq2-basics - Pseudobulk condition DE engine for aggregated counts\n- differential-expression\u002Fde-results - Shrink, filter, and interpret pseudobulk DE results\n- pathway-analysis\u002Fgo-enrichment - Functional interpretation of marker and DE gene lists\n- workflows\u002Fgrn-pipeline - Downstream: infer gene regulatory networks (pySCENIC Path A) from the annotated object\n- workflows\u002Fspatial-pipeline - Downstream: the annotated reference deconvolves spatial spots\n",{"name":5,"source":24,"license":15,"workflow":25,"tool_type":26,"depends_on":27,"description":7,"imported_at":33,"source_path":34,"primary_tool":35,"qc_checkpoints":36,"source_repository":45,"goal_approach_exempt":25},"https:\u002F\u002Fgithub.com\u002FGPTomics\u002FbioSkills",true,"mixed",[28,29,30,31,32],"single-cell\u002Fdata-io","single-cell\u002Fpreprocessing","single-cell\u002Fdoublet-detection","single-cell\u002Fclustering","single-cell\u002Fmarkers-annotation","2026-09-20","workflows\u002Fscrnaseq-pipeline\u002FSKILL.md","Seurat",[37,39,41,43],{"after_loading":38},"Expected cell count, reasonable UMI distribution",{"after_qc":40},"Remove low-quality cells and doublets",{"after_normalization":42},"No batch effects, HVGs look sensible",{"after_clustering":44},"Clusters are biologically meaningful","GPTomics\u002FbioSkills","main","2d1750717747097b602acf30bd09a7cbd4b4c6ea","awesome_skills\u002Fbiology\u002Fbio-workflows-scrnaseq-pipeline","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous","2026-11-09 11:45:36",{"source":8,"skillName":5,"commands":53,"prompt":63,"zipAvailable":25},[54,57,60],{"tool":55,"command":56},"npx","npx skills add Mr-Nobody-Anonymous\u002FAll-skills --skill bio-workflows-scrnaseq-pipeline",{"tool":58,"command":59},"bunx","bunx skills add Mr-Nobody-Anonymous\u002FAll-skills --skill bio-workflows-scrnaseq-pipeline",{"tool":61,"command":62},"pnpm","pnpm dlx skills add Mr-Nobody-Anonymous\u002FAll-skills --skill bio-workflows-scrnaseq-pipeline","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fbiology\u002Fbio-workflows-scrnaseq-pipeline\u002FSKILL.md",[65,71,77],{"path":66,"resourceType":67,"resourceTypeDesc":68,"sizeBytes":69,"sourceHtmlUrl":70},"examples\u002Fscanpy_workflow.py","SCRIPT","脚本",5618,"https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fbiology\u002Fbio-workflows-scrnaseq-pipeline\u002Fexamples\u002Fscanpy_workflow.py",{"path":72,"resourceType":73,"resourceTypeDesc":74,"sizeBytes":75,"sourceHtmlUrl":76},"examples\u002Fseurat_workflow.R","OTHER","其他",5532,"https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fbiology\u002Fbio-workflows-scrnaseq-pipeline\u002Fexamples\u002Fseurat_workflow.R",{"path":78,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":81,"sourceHtmlUrl":82},"usage-guide.md","REFERENCE","参考文档",4155,"https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fbiology\u002Fbio-workflows-scrnaseq-pipeline\u002Fusage-guide.md",[84,92,99,106,113,120],{"skillNo":85,"name":86,"nameZh":6,"description":87,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":88,"categories":89,"keywords":90,"publishedAt":91},"01M7FFFZPC2025JQQHPHDX154J","dei-assessment","Assess diversity, equity, and inclusion with metrics. TRIGGERS - Use when user needs help with dei-assessment related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fdei-assessment\u002FSKILL.md",[],[],"2026-11-16 10:05:34",{"skillNo":93,"name":94,"nameZh":6,"description":95,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":96,"categories":97,"keywords":98,"publishedAt":91},"01M7FFFWN8H4WA7A6Y398QKB6J","dei-action-plan","Create DEI action plans with goals and initiatives. TRIGGERS - Use when user needs help with dei-action-plan related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fdei-action-plan\u002FSKILL.md",[],[],{"skillNo":100,"name":101,"nameZh":6,"description":102,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":103,"categories":104,"keywords":105,"publishedAt":91},"01M7FFFSHTM0QZRVZ2NRJTJ0PY","culture-fit-assessment","Design culture add assessment processes. TRIGGERS - Use when user needs help with culture-fit-assessment related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fculture-fit-assessment\u002FSKILL.md",[],[],{"skillNo":107,"name":108,"nameZh":6,"description":109,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":110,"categories":111,"keywords":112,"publishedAt":91},"01M7FFFMA12RKEQA38KHJSXXPS","cowork-to-code-bridge","Use an already-installed, independently verified cowork-to-code bridge to run narrowly approved actions on the user's own macOS, Linux, or WSL2 machine through a local file queue.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fcowork-to-code-bridge\u002FSKILL.md",[],[],{"skillNo":114,"name":115,"nameZh":6,"description":116,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":117,"categories":118,"keywords":119,"publishedAt":91},"01M7FFFH1F6M8YRGW5V487GH1K","counter-offer-response","Draft counter-offer responses for retention. TRIGGERS - Use when user needs help with counter-offer-response related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fcounter-offer-response\u002FSKILL.md",[],[],{"skillNo":121,"name":122,"nameZh":6,"description":123,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":124,"categories":125,"keywords":126,"publishedAt":91},"01M7FFFDGEHV8Y7J9N40HZXW52","continuous-feedback-system","Design continuous feedback systems. TRIGGERS - Use when user needs help with continuous-feedback-system related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fcontinuous-feedback-system\u002FSKILL.md",[],[]]