RNAseq.SimOptions.2grp: Set up options for simulating RNA-seq data in two-group comparison.

Description

This function takes user provided options for for simulating RNA-seq data, and return a list. The result of this function will be the input for "runSims" and "simRNAseq" function.

Usage

RNAseq.SimOptions.2grp(ngenes, seqDepth, lBaselineExpr, lOD, p.DE, lfc, sim.seed)

Arguments

ngenes

Number of genes in the simulation.

lBaselineExpr

log baseline expression for each gene. This can be: (1) a constant; (2) a vector with length ngenes; (3) a function that takes an integer n, and generate a vector of length n; (4) a string specifying the name of existing datasets, from which the mean expressions will be sampled. Details for this option is provide in "Details" section.

lOD

log over-dispersion for each gene. Available options are the same as for lBaselineExpr.

seqDepth

Sequencing depth, in terms of total read counts. This will be ignored if lBaselineExpr is specified.

p.DE

Percentage of genes being differentially expressed (DE). By default it's 5%.

lfc

log-fold change for DE genes. This can be: (1) a constant; (2) a vector with length being number of DE genes; (3) a function that takes an integer n, and generate a vector of length n. If the input is a vector and the length is not the number of DE genes, it will be sampled with replacement to generate log-fold change.

sim.seed

Simulation seed.

Value

ngenes: An integer for number of genes.
p.DE: Percentage of DE genes.
lBaselineExpr: A vector of length ngenes for log baseline expression.
lOD: A vector of length ngenes for log over-dispersion.
lfc: A vector of length (ngenes*p.DE) for log fold change of the DE genes.
sim.seed: The specified simulation seed.
design: A string representing the experimental design. From this function it is '2grp', standing for two-group comparison.

Details

The simulation of RNA-seq data requires a lot of parameters. This function provides users an interface to specify the simulation parameters. The result from this function will be used for simulating RNA-seq count data. By default, the simulation parameters are similar to that from Cheung data (for unrelated individuals, with large biological variance).

The baseline expression levels and log over-dispersions can be sampled from real data. There are parameters estimated from several real datasets distributed with the package. Available string options for "lBaselineExpr" and "lOD" include: (1) "cheung": parameters from Cheung data, which measures unrelated individuals, so the dispersions are large; (2) "gilad": from Gilad data which are for Human liver sample comparisons between male and female. This dataset has moderate dispersions; (3) "bottomly": from Bottmly data which are from comparing two strains of inbred mice. The dispersions are small. (4) "maqc": from MAQC data which are technical replicates. There are no biological variation from the replicates because the data are technical replicates. The dispersions from this dataset is very small.

The effect sizes (log fold changes of the DE genes) are arbitrarily specified. It is possible to estimate those from real data. We provide a simple example in the package vignette for doing so.

Examples

Run this code

## default
simOptions=RNAseq.SimOptions.2grp()
summary(simOptions)

## specify some parameters: generate baseline expression and
## dispersion from Bottom data, and specify a function for
## alternative log fold changes.
fun.lfc=function(x) rnorm(x, mean=0, sd=1.5)
simOptions=RNAseq.SimOptions.2grp(ngenes=30000, lBaselineExpr="bottomly",
       lOD="bottomly", p.DE=0.05, lfc=fun.lfc)
summary(simOptions)