Trajectory inference — backend zoo

This zoo holds one tutorial per ov.single.TrajInfer(method=...) backend. Every tutorial follows the same template — load → TrajInfer → pseudotime plot → downstream — so you can swap methods by changing one line.

Vendored backends (ship with omicverse)

Method	Tutorial	Strength
Palantir	t_traj_palantir	Branch probabilities; best continuous-differentiation hierarchy.
Diffusion-map / DPT	t_traj_diffusion	scanpy’s diffmap + PAGA.
scTour	t_traj_sctour	Generative; yields a latent velocity vector field.
StaVIA	t_traj_stavia · toy	Spatial-aware variant.
Monocle 2	t_traj_monocle2	DDRTree-based hierarchy.
VIA	t_via · t_via_velo	Markov-chain pseudotime; velocity-aware.
CytoTrace 2	t_cytotrace2	Differentiation potential.

dynbenchmark zoo (install via pip install omicverse[trajectory])

Method	Tutorial	Topology	Backend
SCORPIUS	t_traj_scorpius	linear	pyscorpius
TSCAN	t_traj_tscan	tree	pytscan
destiny	t_traj_destiny	linear, branching	pydestiny-bio
URD	t_traj_urd	branching	pyurd-bio
Monocle 3	t_traj_monocle3	tree	pymonocle3-bio
CytoTRACE	t_traj_cytotrace_bio	gradient	pycytotrace-bio

The Python ports preserve byte-equivalent geodesic distances against the R upstream (dynverse/dyneval), with metric drift |Δ| < 0.07 across the 13-method × 14-dataset benchmark.

For Slingshot (the recommended outer-level backend), see ../t_traj_slingshot.