NBsimSingleCell: Simulate a scRNA-seq experiment
In statOmics/zingeR: Zero-Inflated Negative binomial Gene Expression in R

Description Usage Arguments See Also Examples

View source: R/simulation.R

This function simulates a count matrix for a scRNA-seq experiment based on parameters estimated from a real scRNA-seq count matrix. Global patterns of zero abundance as well as feature-specific mean-variance relationships are retained in the simulation.

NBsimSingleCell(dataset, group, nTags = 10000, nlibs = length(group),
  lib.size = NULL, drop.extreme.dispersion = FALSE, pUp = 0.5,
  foldDiff = 3, verbose = TRUE, ind = NULL, params = NULL,
  randomZero = 0, max.dispersion = 400, min.dispersion = 0.01)

`dataset`	An expression matrix representing the dataset on which the simulation is based.
`group`	Group indicator specifying the attribution of the samples to the different conditions of interest that are being simulated.
`nTags`	The number of features (genes) to simulate. $1000$ by default
`nlibs`	The number of samples to simulate. Defaults to `length(group)`.
`lib.size`	The library sizes for the simulated samples. If `NULL` (default), library sizes are resampled from the original datset.
`drop.extreme.dispersion`	Only applicable if `params=NULL` and used as input to `getDatasetZTNB`. Numeric value between $0$ and $1$ specifying the fraction of highest dispersion values to remove after estimating feature-wise parameters.
`pUp`	Numeric value between $0$ and $1$ ($0.5$ by default) specifying the proportion of differentially expressed genes that show an upregulation in the second condition.
`foldDiff`	The fold changes used in simulating the differentially expressed genes. Either one numeric value for specifying the same fold change for all DE genes, or a vector of the same length as `ind` to specify fold changes for all differentially expressed genes. Note that fold changes above $1$ should be used as input of which a fraction will be inversed (i.e. simulation downregulation) according to 'pUp'. Defaults to $3$.
`verbose`	Logical, stating whether progress be printed.
`ind`	Integer vector specifying the rows of the count matrix that represent differential features.
`params`	An object containing feature-wise parameters used for simulation as created by `getDatasetZTNB`. If `NULL`, parameters are estimated from the dataset provided.
`randomZero`	A numeric value between $0$ and $1$ specifying the random fraction of cells that are set to zero after simulating the expression count matrix. Defaults to $0$.
`min.dispersion`	The minimum dispersion value to use for simulation. $0.1$ by default.
`max.dipserion`	The maximum dispersion value to use for simulation. $400$ by default.

getDatasetZTNB

data(islamEset,package="zingeR")
islam=exprs(islamEset)[1:2000,]
design=model.matrix(~pData(islamEset)[,1])
gamlss.tr::gen.trun(par=0, family="NBI", name="ZeroTruncated", type="left", varying=FALSE)
params = getDatasetZTNB(counts=islam, design=design)
nSamples=80
grp=as.factor(rep(0:1, each = nSamples/2)) #two-group comparison
nTags=2000 #nr of features
set.seed(436)
DEind = sample(1:nTags,floor(nTags*.1),replace=FALSE) #10% differentially expressed
fcSim=(2 + rexp(length(DEind), rate = 1/2)) #fold changes
libSizes=sample(colSums(islam),nSamples,replace=TRUE) #library sizes
simDataIslam <- NBsimSingleCell(foldDiff=fcSim, ind=DEind, dataset=islam, nTags=nTags, group=grp, verbose=TRUE, params=params, lib.size=libSizes)