Slingshot: cell lineage and pseudotime inference for single-cell transcriptomics

BMC Genomics

Table 1 Summary of existing lineage and pseudotime inference methods

	Dimensionality reduction	Cluster based	Graph	Pseudotime calculation	Branching	Supervision
Diffusion Pseudotime	Diffusion maps	No	Weighted k-NN graph on cells	Transition probabilities over arbitrary length random walks	Yes	Starting cell
Embeddr	Laplacian eigenmaps	No	N/A	Principal curve, orthogonal projection	No	Path direction¹, subsetting²
Monocle	ICA	No	MST on cells	Diameter path, PQ trees	Yes³	Path direction¹, number of lineages
Monocle 2	Reversed graph embedding	No	Principal graph on cells	Distance to root	Yes	Starting cluster
TSCAN	PCA	Yes	MST on clusters	Cluster centers, orthogonal projection	Yes	Starting cluster
Waterfall	PCA	Yes	MST on clusters	Cluster centers, orthogonal projection	Yes⁴	Path direction¹
Wishbone	Diffusion maps	No	Ensemble of k-NN graphs on cells	Distance refinement by waypoints	Yes⁵	Starting cell
Slingshot	Any	Yes	MST on clusters	Simultaneous principal curves, orthogonal projection	Yes	Starting cluster, end clusters (optional)

¹Some methods infer a single path or backbone and rely on the user to assign its directionality
²Methods that do not detect branching events require manually subsetting the data down to a single lineage
³Monocle does not detect the number of branching events, the number of lineages must be supplied by the user
⁴Waterfall detects branching events, but requires subsetting to a single lineage for pseudotime calculation
⁵Wishbone can only detect a single branching event (two lineages)

ISSN: 1471-2164