geneSmoothFit: Gene Smooth: Fit expression of genes

In farrellja/URD: URD

Description

This takes a group of genes and cells, averages gene expression in groups of cells (determined using a moving window through pseudotime), and then uses smoothing algorithms (e.g. LOESS or spline fitting) to describe the expression of each gene. The results are returned for use in plotSmoothFit to visualize the fits or several sets of results (e.g. from different trajectories) can be combined with cropSmoothFit and plotted with plotSmoothFitMultiCascade.

Usage

geneSmoothFit(
  object,
  pseudotime,
  cells,
  genes,
  method = c("lowess", "spline"),
  moving.window = 3,
  cells.per.window = NULL,
  pseudotime.per.window = NULL,
  verbose = T,
  verbose.genes = F,
  ...
)

Arguments

object	An URD object
pseudotime	(Character) Name of pseudotime (i.e. a column name of @pseudotime)
cells	(Character vector) Cells to include
genes	(Character vector) Genes to include
method	(Character) Which smoothing method to use ("lowess" calls stats::lowess and "spline" calls stats::smooth.spline)
moving.window	(Numeric) Number of bins to use per window
cells.per.window	(Numeric or NULL) Size of bins (number of cells)
pseudotime.per.window	(Numeric or NULL) Size of bins (pseudotime)
verbose	(Logical) Display progress reports?
verbose.genes	(Logical) Display a message when starting each gene?

Value

(Named List):

• pt.windows: (List) Character vectors of cell IDs of cells in each window, named by either the mean, min, or max pseudotime of those cells (depending on name.by)

• mean.expression: (data.frame) Mean expression of each gene in each pseudotime window

• scaled.expression: (data.frame) Mean expression of each gene in each pseudotime window, scaled to the maximum observed expression

• mean.smooth: (data.frame) Value of fit curve for each pseudotime window (i.e. to predict value of mean.expression)

• scaled.smooth: (data.frame) Value of fit curve for each pseudotime window, scaled to max expression (i.e. to predict value of scaled.expression)

• mean.expression.red: (data.frame) Mean expression reduced to the same dimensions of mean.smooth. (When method is 'spline' if two windows have the same pseudotime, only a single prediction is made for that pseudotime, so this will average the two input windows.)

• scaled.expression.red: (data.frame) Scaled expression reduced to the same dimensions of scaled.smooth.

• method: (Character) Identify fit method ("impulse")

Examples

# Fit a gene expression curve using "spline" method for
# genes "endo.genes" and cells in segments 47 and 46 in the tree.
tentacle.spline <- geneSmoothFit(urd.object, method = "spline", 
  pseudotime = "pseudotime", 
  cells = cellsInCluster(urd.object, "segment", c("47", "46")), 
  genes = endo.genes, moving.window = 1, cells.per.window = 5, spar = 0.9
)

farrellja/URD documentation built on June 17, 2020, 4:48 a.m.