py-Slingshot

A Python port of slingshot (Street et al. BMC Genomics 2018) — branching cell-lineage and pseudotime inference for single-cell RNA-seq.

• Pure NumPy / SciPy / statsmodels implementation
• MST over cluster centroids → DFS lineage enumeration → per-lineage principal curve fitting
• Parity vs R Slingshot on a Y-shaped fixture: pseudotime Spearman 0.99+ along matched lineages

Install

pip install pyslingshot-bio

(module name is pyslingshot; the PyPI distribution name is pyslingshot-bio.)

Quick-start

import pyslingshot
sr = pyslingshot.slingshot(reduced_dim, cluster_labels, start_cluster="root_cluster_id")
# Per-lineage pseudotime
pt = sr.pseudotime              # (n_cells × n_lineages)
# Visualisation
from ggplot2_py import ggsave
ggsave("out.png", plot=pyslingshot.plot_slingshot(sr), width=6, height=4, dpi=120)

Function map

Known limitations (v0.1)

1. Simultaneous principal-curve fitting deferred — v0.1 fits each lineage independently. R Slingshot shares the early stem across lineages during fitting, which produces cleaner branching. v0.2 will add this.
2. MST distance metric uses plain Euclidean on centroids; R uses Mahalanobis-style scaling by within-cluster dispersion when dist.method = "slingshot" (default).
3. embedCurves, predict, branchID deferred to v0.2.
4. Lineage count can differ from R because of root-selection and lineage-merging differences; the matched lineages have Spearman ≥ 0.99 pseudotime parity.

Citation

Street, K. et al. Slingshot: cell lineage and pseudotime inference for single-cell transcriptomics. BMC Genomics 19, 477 (2018).

License

MIT.