Skip to contents

Generates synthetic bulk RNAseq data

synthetic_bulk_cor_matrix.Rd

Function generates synthetic bulkRNAseq data with heteroskedasticity (lowly expressed genes show higher variance) and optional co-expression modules for testing module detection methods. The ground truth comes back with the counts, so a recovered module or eigengene can be scored against what was actually simulated. Which generator to pick and what each parameter does is documented in params_synthetic_bulk_rnaseq().

Usage

synthetic_bulk_cor_matrix(synthetic_params = params_synthetic_bulk_rnaseq())

Arguments

synthetic_params

List. The synthetic data parameters, see params_synthetic_bulk_rnaseq().

Value

A synthetic_bulk_data class containing:

  • counts - The count matrix. Rows are genes, columns are samples.

  • sparse_counts - A slot for sparse counts that can be added later, see simulate_dropouts().

  • module_data - Per-gene ground truth: the gene identifier, its membership (0 for background), its loading on the module factor and whether it is a hub gene.

  • module_factors - The latent factors the modules were built on. Rows are modules, columns are samples. This is what a module eigengene or an ICA/NMF component is trying to recover.

The parameters used are stored on the synthetic_params attribute.

Examples

# three co-expression modules on a 1000 x 100 count matrix
syn <- synthetic_bulk_cor_matrix(params_synthetic_bulk_rnaseq())
dim(syn$counts)
#> [1] 1000  100
head(syn$module_data)
#>      gene membership   loading is_hub
#>    <char>      <int>     <num> <lgcl>
#> 1: gene_1          1 0.7308258  FALSE
#> 2: gene_2          1 5.5270005   TRUE
#> 3: gene_3          1 2.6595206   TRUE
#> 4: gene_4          1 1.0970747  FALSE
#> 5: gene_5          1 1.3758719  FALSE
#> 6: gene_6          1 0.6133037  FALSE
dim(syn$module_factors)
#> [1]   3 100