scde (version 2.0.1)

scde.posteriors: Calculate joint expression magnitude posteriors across a set of cells

Description

Calculates expression magnitude posteriors for the individual cells, and then uses bootstrap resampling to calculate a joint expression posterior for all the specified cells. Alternatively during batch-effect correction procedure, the joint posterior can be calculated for a random composition of cells of different groups (see batch and composition parameters).

Usage

scde.posteriors(models, counts, prior, n.randomizations = 100, batch = NULL,
  composition = NULL, return.individual.posteriors = FALSE,
  return.individual.posterior.modes = FALSE, ensemble.posterior = FALSE,
  n.cores = 20)

Arguments

models

models models determined by scde.error.models

counts

read count matrix

prior

gene expression prior as determined by scde.expression.prior

n.randomizations

number of bootstrap iterations to perform

batch

a factor describing which batch group each cell (i.e. each row of models matrix) belongs to

composition

a vector describing the batch composition of a group to be sampled

return.individual.posteriors

whether expression posteriors of each cell should be returned

return.individual.posterior.modes

whether modes of expression posteriors of each cell should be returned

ensemble.posterior

Boolean of whether to calculate the ensemble posterior (sum of individual posteriors) instead of a joint (product) posterior. (default: FALSE)

n.cores

number of cores to utilize

Value

• default{ a posterior probability matrix, with rows corresponding to genes, and columns to expression levels (as defined by prior$x) } return.individual.posterior.modes{ a list is returned, with the $jp slot giving the joint posterior matrix, as described above. The $modes slot gives a matrix of individual expression posterior mode values on log scale (rows - genes, columns -cells)} return.individual.posteriors{ a list is returned, with the $post slot giving a list of individual posterior matrices, in a form analogous to the joint posterior matrix, but reported on log scale }

Examples

data(es.mef.small)
cd <- clean.counts(es.mef.small, min.lib.size=1000, min.reads = 1, min.detected = 1)
data(o.ifm)  # Load precomputed model. Use ?scde.error.models to see how o.ifm was generated
o.prior <- scde.expression.prior(models = o.ifm, counts = cd, length.out = 400, show.plot = FALSE)
# calculate joint posteriors
jp <- scde.posteriors(o.ifm, cd, o.prior, n.cores = 1)