Table 1.

A summary table for all discussed methods

Method name	Category	Input	Output	Suitability / assumptions	Implementation / access	Software link	Ref.
sceb	ED	1) Budget 2) Pilot data	1) Number of cells to profile 2) Sequencing depth	Determining number of cells and read coverage for any single-cell experiment, including time-series and snapshot studies	Python / Open	https://github.com/martinjzhang/single_cell_eb	69
howmanycells	ED	1) Expected number of cell types 2) Minimum fraction of rarest cell type 3) Minimum number of expected cells per type	1) Number of cells to profile	Determining the number of cells to profile for studies in which rare cell types are either of interest or expected to be important.	HTML JavaScript / Open	https://satijalab.org/howmanycells/	NA
TPS	ED	1) Pilot data 2) A level of reconstruction error	1) The best time points to profile	Cases in which the sampling rate is expected to be able to recover all major molecular events occurring in the process being studied.	Python / Open	http://sb.cs.cmu.edu/TPS/	70
Monocle	TI	Cells by genes matrix	1) Clusters 2) Trajectory graph 3) Pseudotime for all cells	Pseudotime inference for time-series or single-time point (unsynchronized) studies in which the profiled cells are expected to span the entire duration of the process. Clustering can be used for cases in which it is not clear if the sampling rate covers all major biological transitions. Time-series information is not utilized by the method, and so inference is based on the expression levels only.	R / Open	https://cole-trapnell-lab.github.io/monocle3/	37
TSCAN	TI	Cells by genes matrix	1) Clusters 2) Trajectory graph 3) Pseudotime for all cells	Similar to Monocle in terms of assumptions and suitability.	R / Open	https://github.com/zji90/TSCAN	34
Slingshot	TI	Cells by genes matrix	1) Clusters 2) Trajectory graph 3) Pseudotime for all cells	Similar to Monocle in terms of assumptions and suitability.	R / Open	https://github.com/kstreet13/slingshot	35
SLICE	TI	Cells by genes matrix	1) Clusters 2) Trajectory graph 3) Pseudotime for all cells	Relies on entropy-based analysis and so is most suitable for developmental studies. It can infer the starting set of cells on its own without using the time information.	R / Open	https://github.com/xu-lab/SLICE	39
PAGA	TI	Cells by genes matrix	1) Clusters 2) Trajectory graph 3) Pseudotime for all cells	Suitable for learning complex trajectory structure with multiple branching. Efficient and often fairly fast.	Python / Open	https://github.com/theislab/paga	36
Seurat	TI, IM	Cells by genes matrix	1) Clusters 2) Trajectory graph 3) Pseudotime for all cells	An inclusive suite of tools for the analysis of scRNA-seq data. It provides implementation of several methods. It is very efficient and fast but may be less accurate for complex trajectories.	R / Open	https://satijalab.org/seurat/	53,71
SCUBA	TI	Cells by genes matrix	1) Clusters 2) Trajectory graph 3) Pseudotime for all cells	Does not infer continuous trajectories, only states and their relationships. It is mainly successful for cases where the trajectory is linear or includes few branches.	MATLAB / Open	https://github.com/gcyuan/SCUBA	48
scdiff	TI, GRN IM	Cells by genes matrix	1) Clusters 2) Trajectory graph 3) Pseudotime for all cells 4) Regulatory networks	Does not infer continuous trajectories, only states and their relationships. It is suitable for learning complex branching models. It works in gene space and so is suitable for cases in which several cell types are expected.	Python, JavaScript /Open	https://github.com/phoenixding/scdiff	38
CSHMM	TI, GRN, IM	Cells by genes matrix	1) Clusters 2) Trajectory graph 3) Pseudotime for all cells 4) Regulatory networks	Infers continuous trajectories in gene space and so can generate complex branching models for cases in which several cell types are expected. It is slower than most methods that work on reduced dimension space.	Python / Open	https://github.com/jessica1338/CSHMM-for-time-series-scRNA-Seq	49
RNA velocity and scVelo	TI	scRNA-seq data (all reads)	RNA velocity vectors for all cells	These methods do not rely on expression similarity and so may be better suited for data sampled at intervals that do not fully capture all possible molecular events. As the methods are based on identifying splicing status, they may be problematic for any scRNA-seq datasets where the reads do not sufficiently cover both intronic and exonic regions. They are often a valuable complement to pseudotime inference methods.	Python / Open	https://github.com/theislab/scvelo	9,51,72
LinTIMaT	TI, IM	1) Cells by barcodes matrix 2) Cells by genes matrix	Trajectories and barcodes that mark different cell fates	Assumes the existence of CRISPR-based mutation information for cells. Learns linage models by combining these with scRNA-seq expression data.	Python / Open	https://jessica1338.github.io/LinTIMaT/	57
PhenoPath	TI, IM	1) Cells by genes matrix 2) Vectors of covariates	1) Clusters 2) Trajectory graph 3) pseudotime for all cells	Suitable for cases in which data from multiple individuals, perhaps at different diseases or development stages, are integrated. Does not assume or require a clear ordering of the samples.	R / Open	https://bioconductor.org/packages/release/bioc/html/phenopath.html	60
TimeReg	GRN, IM	1) Expression matrix for RNA-seq data 2) Bam files for epigenetics data	1) Regulatory network (TFs and target genes)	Requires information about TF–gene interactions.	R / Open	https://github.com/SUwonglab/TimeReg	61
GRNVBEM	GRN, IM	1) Cells by genes matrix	1) Regulatory network	Can either use time series or pseudotime-inferred ordering to reconstruct the GRN.	MATLAB / Open	https://github.com/mscastillo/GRNVBEM	62

CSHMM, continuous-state hidden Markov model; ED, experimental design; GRN, gene regulatory network; GRNVBEM, a gene regulatory network (GRN) inference method using a variational Bayesian expectation-maximization (VBEM) framework; IM, integrative model; LinTIMaT, lineage tracing by integrating mutation and transcriptomic data; PAGA, partition-based graph abstraction; RGN, regulatory network inference; RNA-seq, RNA sequencing; sceb, single-cell empirical Bayes; scRNA-seq, single-cell RNA sequencing; SCUBA, single-cell clustering using bifurcation analysis; SLICE, single-cell lineage inference using cell expression similarity and entropy; TF, transcription factor; TI, trajectory inference; TPS, time point selection; TSCAN, tools for single-cell analysis.