Node reference

CellRank nodes.

CellRank fate analysis as a Markov chain over cells: build a transition kernel, identify macrostates and terminal / initial states with GPCCA, compute fate probabilities, find the genes that drive each lineage, and draw GAM-smoothed gene trends along them.

05 nodesCategory · cellrankin 3 templates

Typical pipeline

Neighbors → CellRank Kernel (CytoTRACE / Pseudotime / Velocity / Connectivity) → Cell Fate Analysis → Lineage Drivers → CellRank Visualization and Gene Trends. The CytoTRACE and connectivity kernels work without RNA velocity; the pseudotime kernel wants DPT upstream; the velocity kernel wants RNA Velocity. Three shipped templates wire these up.

CellRank Kernel

cellrankKernel

Compute cell-cell transition probabilities (Markov chain). Kernel: cytotrace / pseudotime / velocity / connectivity.

Node LibraryscRNA-seq

InAnnData (with neighbors)

OutAnnData (obsp["T_fwd"])

Parameters (5)

Name	Type	Default	Description
`kernel_type`	select	`cytotrace`	cytotrace/pseudotime/velocity/connectivity.
`time_key`	obs column	`dpt_pseudotime`	Pseudotime obs column (pseudotime kernel).
`threshold_scheme`	select	`hard`	Threshold scheme for pseudotime kernel: 'hard' (binary) or 'soft' (logistic). Only used when kernel_type='pseudotime'.
`use_connectivity_mix`	checkbox	`true`	Mix with ConnectivityKernel.
`weight`	number	`0.2`	Connectivity mixing weight.

Cell Fate Analysis

cellFate

GPCCA — identify terminal/initial macrostates and compute fate probabilities.

Node LibraryscRNA-seq

InAnnData (with kernel)

OutAnnData (obsm fate probabilities, obs macrostates)

Parameters (7)

Name	Type	Default	Description
`n_states`	number	—	null = auto.
`cluster_key`	obs column	—	Cluster column to guide macrostates.
`terminal_method`	select	`stability`	stability/eigengap/top_n.
`n_terminal_states`	number	—	Number of terminal states to predict. Only used when terminal_method='top_n'; ignored by 'stability'/'eigengap' (which determine count automatically).
`fit_initial_states`	checkbox	`true`	Also predict initial states.
`compute_fate_probs`	checkbox	`true`	Compute fate probabilities.
`solver`	select	`gmres`	Linear solver for fate probabilities: gmres (fast), direct (exact), lgmres, bicgstab. Only used when compute_fate_probs=true; ignored when fate probabilities are skipped.

Lineage Drivers

lineageDrivers

Identify genes driving differentiation toward each cell fate (correlation with fate probabilities). Unlike DEG (cluster vs cluster), captures genes that change continuously along trajectories.

Node LibraryscRNA-seq

InAnnData (with fate probabilities)

OutAnnData (uns["lineage_drivers"], uns["lineage_drivers_df"])

Parameters (5)

Name	Type	Default	Description
`lineages`	text	—	Comma-separated lineage names.
`n_top_genes`	number	`50`	Top driver genes per lineage.
`cluster_key`	obs column	—	Restrict to specific clusters.
`clusters`	text	—	Comma-separated groups of cluster_key to restrict the correlation to (empty = all cells).
`use_raw`	checkbox	`true`	Correlate fate probabilities with all log-normalized genes in adata.raw instead of only the genes in X.

CellRank Visualization

cellrankVisualization

Plot fate analysis results (fate probs, macrostates, terminal/initial states, circular projection, coarse transition, lineage drivers).

Node LibraryscRNA-seq

InAnnData (with fate analysis)

OutPNG image

Plots (7)fate_probabilitiesmacrostatesterminal_statesinitial_statescircular_projectioncoarse_transitionlineage_drivers

Parameters (4)

Name	Type	Default	Description
`plot_type`	text	`fate_probabilities`	7 plot types.
`embedding`	text	`umap`	umap/tsne/diffmap.
`lineage`	text	—	Specific lineage(s), comma-separated.
`n_top_genes`	number	`15`	Top driver genes (lineage_drivers only).

Gene Trends

geneTrends

GAM-smoothed expression trends along differentiation lineages. Auto-selects from Lineage Drivers if Genes is empty. Two modes: trends (smoothed curves) or heatmap.

Node LibraryscRNA-seq

InAnnData (with fate probabilities)

OutPNG image

Plots (2)trendsheatmap

Parameters (4)

Name	Type	Default	Description
`genes`	text	—	Comma-separated genes (empty = auto).
`lineages`	text	—	Comma-separated lineages.
`plot_type`	text	`trends`	"trends" or "heatmap".
`n_top_genes`	number	`10`	Auto-selected gene count.

Trajectory

Annotation