AssignPhase: Classify Cells into Cell Cycle Stages

In tallulandrews/CycleMix: Cell-Cycle using Mixture Models

Classify Cells into Cell Cycle Stages

Description

Uses mixed-gaussian distributions to assign cells to different cell-cycle stages according to provided marker/DE genes.

Usage

  assignPhase(SCE, CC_table, phase="G2M", expr_name="logcounts", do.scale=FALSE, symbol_column="feature_symbol")
  classifyCells(SCE, CC_table, expr_name="logcounts", do.scale=FALSE, symbol_column="feature_symbol", allow.multi=FALSE)

Arguments

SCE	a SingleCellExperiment object.
CC_table	a dataframe with three columns: Gene, Stage, Dir
phase	which cell-cycle stage to assign cells to, must be a value found in "Stage" of CC_table
expr_name	name of the expression matrix to use, must be an element of assays(SCE), should be at minimum library-size normalized and log-transformed.
do.scale	mean-center and scale variance of each gene.
symbol_column	name of the column of rowData(SCE) which contains HGNC gene symbols, if null uses rownames.
allow.multi	allow cells to be assigned to multiple cell-cycle phases.

Details

Classifies cells as being in a particular phase of the cell-cycle by modelling the expression signature of each stage as a mixture of guassian distributions. The gene expression signature of a particular stage as the weighted mean expression of the stage-specific genes provided in the CC_table Weights are provided in the "Dir" column of the CC_table.

Each stage specific signature is fit as a mixture of 1-3 guassian distributions and BIC is used to select the best model using mclust. If the best model is a mixture of >1 distributions then cells assigned to the distribution with the highest mean are assigned to the respective cell-cycle phase.

If a cell is assigned to multiple cell-cycle phases and allow.multi is FALSE then signatures for each phase are centered and scaled, cells are assigned to the phase for which it has the highest signature.

assignPhase performs the fitting and assigns cells to a single cell-cycle phase classifyCells runs assignPhase for all phases in the CC_table and combines the output into a single assignment vector.

Value

assignPhase A list of output from mclust with a extra element, "phase", which is a vector of phase assignments for each cell. classifyCells A list of three elements:

"phase" : a vector of phase assignments for each cell.

"scores" : a table of expression signatures for each cell to each phase.

"fits" : a list of output from assignPhase.

Examples

	fit <- assignPhase(Ex, MGeneSets$Cyclone, phase="G2M")
	classification <- classifyCells(Ex,  MGeneSets$Cyclone, do.scale=TRUE, allow.multi=FALSE);
	classification <- classifyCells(Ex,  MGeneSets$Cyclone, do.scale=FALSE, allow.multi=TRUE);

tallulandrews/CycleMix documentation built on Sept. 28, 2022, 6:55 p.m.