Defines functions dimReduce

Documented in dimReduce

#' Get lower dimension embedding
#' @param x    gene expresison matrix [GENES x SAMPLES]
#' @param flavor    the algorithm to use to obtain the dimensionality reduction
#'                  must be in c('pca', 'tsne', 'umap')
#' @param k    the number of dimensions in the reduced dimension representation
#' @param is.counts    logical: is `x` counts data
#' @param ntop    number of most variable genes to use for dimensionality
#'                reduction
#' @return    reduced dimensionality representation
#' @keywords internal
#' @importFrom scater runPCA runTSNE runUMAP calculateCPM
#' @importFrom SingleCellExperiment reducedDim
#' @import SingleCellExperiment
dimReduce <- function(x, flavor=c('pca', 'tsne', 'umap'), k=2, is.counts=TRUE, ntop=500) {
  flavor <- match.arg(flavor)
  if(flavor=='pca') function.to.call <- runPCA
  if(flavor=='tsne') function.to.call <- runTSNE
  if(flavor=='umap') function.to.call <- runUMAP

    sce <- SingleCellExperiment(assays=list(counts=x))
    exprs(sce) <- log2(calculateCPM(sce) + 1)
  } else {
    sce <- SingleCellExperiment(assays=list(logcounts=x))

  sce <- function.to.call(sce, ncomponents=k, ntop=ntop)
  red.dim <- data.frame(reducedDim(sce))[,seq_len(k)]
  colnames(red.dim) <- paste0(flavor, seq_len(k))

Try the netSmooth package in your browser

Any scripts or data that you put into this service are public.

netSmooth documentation built on Nov. 8, 2020, 5:33 p.m.