sc-pseudotime

sc-pseudotime

Use from a step

trajectory = load_skill("sc-pseudotime")
result = trajectory.pseudotime(read_input("clustered.h5ad"), cluster_key="leiden",
                                use_rep="X_pca", root_cluster="0")
write_output(trajectory.pseudotime_table(result), "tables/pseudotime.csv")
write_output(trajectory.trajectory_genes(result), "tables/trajectory_genes.csv")
write_output(trajectory.pseudotime_figure(result), "figures/pseudotime.png")
write_output(result, "intermediate/adata_pseudotime.h5ad")

The input is not changed. The PBMC example in examples/example_step.py demonstrates the API; PBMC labels do not validate a differentiation trajectory.

API

pseudotime(adata, *, method: str='dpt', cluster_key: str='leiden', use_rep: str | None=None, root_cluster: str | None=None, root_cell: str | int | None=None, end_clusters: list[str] | None=None, n_neighbors: int=15, n_pcs: int=50, n_dcs: int=10, palantir_knn: int=30, palantir_n_components: int=10, palantir_num_waypoints: int=1200, palantir_max_iterations: int=25, palantir_seed: int | None=None, via_knn: int=30, via_seed: int | None=None, cellrank_n_states: int=3, cellrank_schur_components: int=20, cellrank_frac_to_keep: float=0.3, cellrank_use_velocity: bool=False, random_state: int=20)

Return a copy with pseudotime and backend diagnostics.

The default representation prefers X_pca; UMAP remains the preferred display embedding, not the default inference graph. This function does not disable numba JIT. Roots express an analysis assumption, not a direction inferred from the pseudotime values.

:param method: dpt, palantir, via, cellrank, slingshot_r or monocle3_r. :param cluster_key: Existing obs grouping column; default leiden. :param use_rep: Existing obsm representation; None prefers X_pca. :param root_cluster: Optional root group for the selected backend. :param root_cell: Optional obs name or integer position. :param end_clusters: Optional Slingshot terminal groups. :param n_neighbors: DPT/CellRank neighbors; default 15. :param n_pcs: PCs for those neighbors; default 50. :param n_dcs: Diffusion components used for DPT; default 10. :param palantir_knn: Palantir neighbors, default 30. :param palantir_n_components: Palantir diffusion components, default 10. :param palantir_num_waypoints: Palantir waypoints, default 1200. :param palantir_max_iterations: Palantir iterations, default 25. :param palantir_seed: Overrides random_state for Palantir; default None. :param via_knn: VIA neighbors, default 30. :param via_seed: Overrides random_state for VIA; default None. :param cellrank_n_states: CellRank states, default 3. :param cellrank_schur_components: CellRank Schur components, default 20. :param cellrank_frac_to_keep: CellRank state-cell fraction, default 0.3. :param cellrank_use_velocity: Couple CellRank to existing velocity; default False. :param random_state: Palantir/VIA seed, default 20. DPT retains Scanpy's deterministic defaults; R wrappers do not expose a seed. :returns: New AnnData with obs['pseudotime']; run_info names the backend and original pseudotime column. R curves remain available separately. :raises ValueError: The method, grouping, expression contract or embedding is invalid. :raises ImportError: An optional backend is missing.