thin_gene: Binomial thinning for altering total gene expression levels

View source: R/genthin.R

thin_geneR Documentation

Binomial thinning for altering total gene expression levels


Given a matrix of real RNA-seq counts, this function will apply a separate, user-provided thinning factor to each gene. This uniformly lowers the counts for all samples in a gene. The thinning factor should be provided on the log2-scale. This is a specific application of the binomial thinning approach in thin_diff. The method is described in detail in Gerard (2020).


thin_gene(mat, thinlog2, relative = FALSE, type = c("thin", "mult"))



A numeric matrix of RNA-seq counts. The rows index the genes and the columns index the samples.


A vector of numerics. Element i is the amount to thin (on the log2 scale) for gene i. For example, a value of 0 means that we do not thin, a value of 1 means that we thin by a factor of 2, a value of 2 means we thin by a factor of 4, etc.


A logical. Should we apply relative thinning (TRUE) or absolute thinning (FALSE). Only experts should change the default.


Should we apply binomial thinning (type = "thin") or just naive multiplication of the counts (type = "mult"). You should always have this set to "thin".


A list-like S3 object of class ThinData. Components include some or all of the following:


The modified matrix of counts.


The design matrix of variables used to simulate signal. This is made by column-binding design_fixed and the permuted version of design_perm.


A matrix of coefficients corresponding to designmat.


Additional variables that should be included in your design matrix in downstream fittings. This is made by column-binding the vector of 1's with design_obs.


A matrix of estimated surrogate variables. In simulation studies you would probably leave this out and estimate your own surrogate variables.


A matrix of target correlations between the surrogate variables and the permuted variables in the design matrix. This might be different from the target_cor you input because we pass it through fix_cor to ensure positive semi-definiteness of the resulting covariance matrix.


A matrix of simulated variables used to permute design_perm if the target_cor is not NULL.


David Gerard


  • Gerard, D (2020). "Data-based RNA-seq simulations by binomial thinning." BMC Bioinformatics. 21(1), 206. doi: 10.1186/s12859-020-3450-9.

See Also


For subsampling the rows and columns of your real RNA-seq count matrix prior to applying binomial thinning.


For the more general thinning approach.


For thinning sample-wise instead of gene-wise.


For thinning all counts uniformly.


For converting a ThinData object to a SummarizedExperiment object.


For converting a ThinData object to a DESeqDataSet object.


## Generate count data and thinning factors
## In practice, you would obtain mat from a real dataset, not simulate it.
n <- 10
p <- 1000
lambda <- 1000
mat <- matrix(lambda, ncol = n, nrow = p)
thinlog2 <- rexp(n = p, rate = 1)

## Thin total gene expressions
thout <- thin_gene(mat = mat, thinlog2 = thinlog2)

## Compare empirical thinning proportions to specified thinning proportions
empirical_propvec <- rowMeans(thout$mat) / lambda
specified_propvec <- 2 ^ (-thinlog2)
plot(empirical_propvec, specified_propvec,
     xlab = "Empirical Thinning Proportion",
     ylab = "Specified Thinning Proportion")
abline(0, 1, col = 2, lwd = 2)

seqgendiff documentation built on March 18, 2022, 5:21 p.m.