makeInfReps: Make pseudo-inferential replicates from mean and variance

In mikelove/fishpond: Fishpond: downstream methods and tools for expression data

Make pseudo-inferential replicates from mean and variance

Description

Makes pseudo-inferential replicate counts from mean and variance assays. The simulated counts are drawn from a negative binomial distribution, with mu=mean and size set using a method of moments estimator for dispersion.

Usage

makeInfReps(y, numReps, minDisp = 0.001)

Arguments

y	a SummarizedExperiment
numReps	how many inferential replicates
minDisp	the minimal dispersion value, set after method of moments estimation from inferential mean and variance

Details

Note that these simulated counts only reflect marginal variance (one transcript or gene at a time), and do not capture the covariance of counts across transcripts or genes, unlike imported inferential replicate data. Therefore, makeInfReps should not be used with summarizeToGene to create gene-level inferential replicates if inferential replicates were originally created on the transcript level. Instead, import the original inferential replicates.

Value

a SummarizedExperiment

References

Van Buren, S., Sarkar, H., Srivastava, A., Rashid, N.U., Patro, R., Love, M.I. (2020) Compression of quantification uncertainty for scRNA-seq counts. bioRxiv. https://doi.org/10.1101/2020.07.06.189639

Examples

library(SummarizedExperiment)
mean <- matrix(1:4,ncol=2)
variance <- mean
se <- SummarizedExperiment(list(mean=mean, variance=variance))
se <- makeInfReps(se, numReps=50)

mikelove/fishpond documentation built on Aug. 29, 2023, 2:45 p.m.