spatial-trajectory

spatial-trajectory

When to use

Infer pseudotime from log-normalized spatial expression with an existing PCA and neighbors graph. DPT is the default; CellRank adds fate probabilities and Palantir adds waypoint branch probabilities.

Use from a step

from skills._sdk.notebook import load_skill
library = load_skill("spatial-trajectory")
library.trajectory(adata, root_cell=str(adata.obs_names[0]))

Run examples/example_step.py with the step runner for a synthetic, executable example.

API

trajectory(adata, *, method: str='dpt', cluster_key: str | None=None, root_cell: str | None=None, root_cell_type: str | None=None, method_params: dict | None=None, random_state: int=0, palantir_waypoint_seed: int=20)

Infer pseudotime in place from log-normalized X and an existing PCA/graph.

DPT and CellRank add dpt_pseudotime; Palantir adds palantir_pseudotime. CellRank fate calculations can be incomplete; inspect run_info warnings.

:param adata: AnnData with X_pca and a neighbors graph from preprocessing. :param method: dpt (CLI default), cellrank or palantir. :param cluster_key: Observation annotation; None detects leiden/cell_type/cluster. :param root_cell: Starting barcode; None picks the maximum first diffusion component. :param root_cell_type: Restrict automatic root selection to this annotation value. :param method_params: Backend options using the CLI names with underscores; None retains its defaults, such as dpt_n_dcs=10. See references/parameters.md. :param random_state: Seed for diffusion maps and supported backend sampling, default 0. :param palantir_waypoint_seed: Palantir waypoint sampling seed, 20 to preserve the original CLI backend default independently from diffusion-map seed 0. :returns: The same AnnData with pseudotime and JSON run diagnostics. :raises ValueError: Missing preprocessing, root or annotation, or invalid method. :raises ImportError: A backend is missing; use install_skill_deps.