Scoring the cell cycle

cellula implements one method for cell cycle scoring, as done by the tricycle package by Zheng and colleagues.

The assignCellCycle() function is a simple wrapper to tricycle with a couple added features: two methods to determine the origin of the PCA point cloud automatically, and a simple way to infer the cell cycle phase (as a categorical vector) using the authors’ indications.

sce = assignCellCycle(sce)

The methods for recentering (recenter argument) are:

• optim an optimization routine that finds the point within the PCA point cloud that is farthest from all other points (by Pierre-Luc Germain)

• chull finds the center by finding the centroid of the convex hull of the cloud

• none takes (0,0) as center.

Additional parameters feed directly into the tricycle::estimate_cycle_position() function, in particular:

• gname.type the type of gene names (rownames of the SCE object), either "SYMBOL" or "ENSEMBL"

• gname the rowData(sce) column with gene identifiers, if not rownames

• expres_values the assay to use to project data on the reference space.

The plotCellCycle() function takes tricycle results and plots them in a circular embedding in which the position of each cell reflects the cell cycle phase it is estimated to be in.

The rings_by argument controls how many rings the data will be divided into: it can be clusters, samples, batches, etc.

The colour_by argument controls the data used for the color scale. By default it is the position (angle) on the projection, which, being cyclical, requires a cyclical color palette.

# Let's make shorter labels
sce$labels_AUC = gsub(sce$labels_AUC, pattern = "MURARO_PANCREAS_", replacement = "")
sce$labels_AUC = gsub(sce$labels_AUC, pattern = "_CELL", replacement = "")

plotCycle(sce, rings_by = "labels_AUC")