Simulate sequencing-depth dropouts on synthetic bulk data

This function induces Splatter-style sequencing-depth sparsity on the data. Per sample a size factor s_j ~ LogNormal(0, capture_efficiency_sigma) is drawn, giving a target library size of target_library_size * s_j. Each gene is then binomially thinned to approach that target, so dropout falls out of the library size rather than an explicit per-gene dropout curve. Retention probability is capped at 1, meaning samples already below their target are left alone rather than upsampled.

Usage

simulate_dropouts(object, sparsity_params = params_bulk_sparsity())

Arguments

object

The synthetic_bulk_data class.

sparsity_params

List. The sparsification parameters, see params_bulk_sparsity().

Value

synthetic_bulk_data with added sparse data.

References

Zappia, et al., Genome Biol, 2017

Examples

# thin the counts down to a shallower library size
syn <- synthetic_bulk_cor_matrix()
syn <- simulate_dropouts(syn, params_bulk_sparsity())
mean(syn$counts == 0)
#> [1] 0.00433
mean(syn$sparse_counts == 0)
#> [1] 0.01176