菜单
Single-cell pseudotime and lineage inference after clustering, with DPT, Palantir, VIA, CellRank, or Slingshot plus post-hoc trajectory gene ranking.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/mdbabumiamssm/LLMs-Universal-Life-Science-and-Clinical-Skills- --skill sc-pseudotime命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
正在显示 SKILL.md
| name | sc-pseudotime |
| description | Single-cell pseudotime and lineage inference after clustering, with DPT, Palantir, VIA, CellRank, or Slingshot plus post-hoc trajectory gene ranking. |
| version | 0.5.0 |
| author | OmicsClaw |
| license | MIT |
| tags | ["singlecell","pseudotime","trajectory","dpt","palantir","via","cellrank","slingshot"] |
| metadata | {"omicsclaw":{"domain":"singlecell","allowed_extra_flags":["--method","--cluster-key","--use-rep","--root-cluster","--root-cell","--end-clusters","--n-neighbors","--n-pcs","--n-dcs","--n-genes","--corr-method","--palantir-knn","--palantir-n-components","--palantir-num-waypoints","--palantir-max-iterations","--palantir-seed","--via-knn","--via-seed","--cellrank-n-states","--cellrank-schur-components","--cellrank-frac-to-keep","--cellrank-use-velocity","--r-enhanced"],"param_hints":{"dpt":"[Truncated]","palantir":"[Truncated]","via":"[Truncated]","cellrank":"[Truncated]","slingshot_r":"[Truncated]","monocle3_r":"[Truncated]"},"saves_h5ad":true,"requires_preprocessed":true,"trigger_keywords":["pseudotime","trajectory","lineage","diffusion pseudotime","palantir","via","cellrank","monocle3","slingshot"]}} |
sc-pseudotime is the trajectory step after clustering. It takes a normalized, cluster-labeled AnnData object, picks a start state, infers a trajectory, then ranks genes associated with that trajectory.
This skill answers questions like:
sc-preprocessingsc-batch-integrationsc-clusteringsc-cell-annotation if the user wants to define the start state by biology instead of just cluster numberIf the user only says “do pseudotime”, the first thing to explain is:
X must represent normalized_expressionadata.obsX_umap when it is already present, otherwise it falls back to integrated or PCA embeddings such as X_harmony, X_scvi, X_scanvi, X_scanorama, or X_pcalayers["counts"] may still exist, but trajectory-gene ranking is performed on normalized adata.Xdptpalantirviacellrankslingshot_rmonocle3_r| Method | Best first use | What it adds |
|---|---|---|
dpt |
| most users’ first trajectory pass |
| classic diffusion pseudotime after graph construction |
palantir | when entropy / fate probabilities are important | waypoint-based pseudotime and terminal-state probabilities |
via | when users want graph-based terminal-state discovery | fast graph trajectory with branch-aware outputs |
cellrank | when users explicitly want macrostates or fate inference | transition-kernel / fate model on top of a graph |
slingshot_r | when users want explicit branch curves | lineage-centric branch inference through the R bridge |
monocle3_r | when users want principal graph trajectory with branching | Monocle3 principal graph trajectory and pseudotime via R bridge. Produces monocle3_pseudotime in obs and MST trajectory edges. Requires monocle3 R package. |
Note on monocle3_r root selection: monocle3_r auto-selects the root principal node by finding the principal graph node nearest to the centroid of the --root-cluster cells in UMAP space. If --root-cluster is not specified, the largest cluster is used. Pass --root-cluster <name> to control trajectory direction.
--method--cluster-key--use-rep--root-cluster or --root-cell--n-genes--corr-methoddpt
--n-neighbors--n-pcs--n-dcspalantir
--palantir-knn--palantir-n-components--palantir-num-waypoints--palantir-max-iterations--palantir-seedvia
--via-knn--via-seedcellrank
--cellrank-n-states--cellrank-schur-components--cellrank-frac-to-keep--cellrank-use-velocityslingshot_r
--end-clustersmonocle3_r
--root-cluster (auto-selects largest cluster if not specified)# Monocle3 pseudotime demo
python omicsclaw.py run sc-pseudotime --demo --method monocle3_r --output /tmp/monocle3_demo
# Monocle3 pseudotime with user data
python omicsclaw.py run sc-pseudotime --input <processed.h5ad> --method monocle3_r --cluster-key leiden --output <dir>
# Monocle3 with explicit root cluster
python omicsclaw.py run sc-pseudotime --input <processed.h5ad> --method monocle3_r --cluster-key cell_type --root-cluster "HSC" --output <dir>
corr_methodprocessed.h5adfigure_dataSuccessful runs write:
processed.h5adreport.mdresult.jsontables/trajectory_genes.csvtables/pseudotime_cells.csvtables/trajectory_summary.csvfigure_data/figures/Common figures include:
pseudotime_embedding.pngpseudotime_distribution_by_group.pngtrajectory_gene_heatmap.pngtrajectory_gene_trends.pngfate_probability_heatmap.png when availablepaga_graph.png for dptlineage_curves.png for slingshot_rmonocle3_trajectory_graph.png for monocle3_rsc-pathway-scoring for lineage signaturessc-enrichment for statistical pathway interpretation of trajectory genessc-de if the user wants condition-level testing after states are stabilizedmethod and corr_method into one storyFor short guardrails see knowledge_base/knowhows/KH-sc-pseudotime-guardrails.md.
For longer method guidance see knowledge_base/skill-guides/singlecell/sc-pseudotime.md.
| Flag | Type | Default | Description | Validation |
|---|---|---|---|---|
--input | str | — | Input .h5ad file | required unless --demo |
--output | str | — | Output directory | required |
--demo | flag | off | Run with bundled PBMC3k data | — |
--method | str | dpt | Trajectory method: dpt, palantir, via, cellrank, slingshot_r, monocle3_r | validated against METHOD_REGISTRY |
--cluster-key | str | leiden | Cluster/label column in obs | — |
--use-rep | str | None | Embedding to use (e.g., X_pca, X_harmony) | auto-detected when omitted |
--root-cluster | str | None | Cluster name to use as trajectory root | — |
--root-cell | str | None | Specific cell (obs_name or integer index) to use as root | — |
--end-clusters | str | None | Comma-separated terminal cluster names | slingshot_r only |
--n-neighbors | int | 15 | KNN size for graph construction | dpt only |
--n-pcs | int | 50 | PCs to use for diffusion map | dpt only |
--n-dcs | int | 10 | Number of diffusion components | dpt only |
--n-genes | int | 50 | Number of trajectory-correlated genes to rank | — |
--corr-method | str | pearson | Gene-pseudotime correlation method | choices: pearson, spearman |
--palantir-knn | int | 30 | KNN for Palantir graph | palantir only |
--palantir-n-components | int | 10 | Diffusion components for Palantir | palantir only |
--palantir-num-waypoints | int | 1200 | Waypoints for Palantir | palantir only |
--palantir-max-iterations | int | 25 | Max Palantir iterations | palantir only |
--palantir-seed | int | 20 | Palantir random seed | palantir only |
--via-knn | int | 30 | KNN for VIA graph | via only |
--via-seed | int | 20 | VIA random seed | via only |
--cellrank-n-states | int | 3 | Number of macrostates | cellrank only |
--cellrank-schur-components | int | 20 | Schur decomposition components | cellrank only |
--cellrank-frac-to-keep | float | 0.3 | Fraction of cells to keep for macrostate estimation | cellrank only |
--cellrank-use-velocity | flag | off | Use velocity kernel in CellRank | cellrank only |
--r-enhanced | flag | off | Also render R Enhanced ggplot2 figures | — |
Activated by --r-enhanced. Files written to figures/r_enhanced/.
| Renderer | Output file | figure_data CSV | Plot description | Required R packages |
|---|---|---|---|---|
plot_pseudotime_lineage | r_pseudotime_lineage.png | pseudotime_points.csv | Pseudotime lineage scatter with loess trajectory curves | ggplot2 |
plot_pseudotime_dynamic | r_pseudotime_dynamic.png | trajectory_genes.csv | Dynamic gene expression trends along pseudotime | ggplot2 |
plot_pseudotime_heatmap | r_pseudotime_heatmap.png | gene_expression.csv | Heatmap of top trajectory genes ordered by pseudotime | ggplot2, ComplexHeatmap |
plot_embedding_discrete | r_embedding_discrete.png | embedding_points.csv | UMAP/embedding colored by cluster labels | ggplot2 |
plot_embedding_feature | r_embedding_feature.png | embedding_points.csv | UMAP/embedding colored by pseudotime score | ggplot2 |
plot_cell_density | r_cell_density.png | embedding_points.csv | 2D density overlay on embedding | ggplot2, ggdensity |
Upstream: sc-clustering Downstream: sc-velocity (RNA velocity), sc-gene-programs (temporal gene programs)
基于 SOC 职业分类
