An R package for simulating cell-type specific and hierarchical single-cell expression data.The hierarchicell package estimates important parameters from single-cell RNAseq expression counts before simulating expression data that is cell-type specific and hierarchical in nature. With the simulated data, hierarchicell is then able to compute power calculations under a variety of conditions. The package consists of four important categories of functions: data loading and cleaning, empirical estimation of distributions, simulating expression data, and computing type 1 error or power.
The data loading and cleaning function is very basic, but data input is critical to the package working correctly. If no input data is given, the package default data will be used for simulation and power calculations.
The most fundamental component of this package is in the estimation of the simulation parameters. The functions to estimate parameters for the simulation estimate the empirical distributions for library size, dropout rate, and global gene means and model the hierarchical variance structure of the input data.
With the parameters estimated, the package can simulate data under a variety of pre-determined conditions. These conditions include foldchange, number of genes, number of samples (i.e., independent experimental units), and the mean number of cells per individual.Either binary (case/control) or continuous outcomes of interest can be simulated.
With the parameters estimated, the package can compute type 1 error rates for a number of different tools under a variety of pre-determined conditions. These conditions include number of genes, number of samples (i.e., independent experimental units), and the mean number of cells per individual.
With the parameters estimated, the package can compute power under a variety of pre-determined conditions. These conditions include foldchange, number of genes, number of samples (i.e., independent experimental units), and the mean number of cells per individual. Power can be calculated for either binary or continuous outcomes of interest.
To install, type the following command:
devtools::install_github("kdzimm/hierarchicell")
### Load and filter a cell-type specific scRNA-seq datset of interest
# (Here, we are just loading the default dataset)
clean_expr_data <- filter_counts()
### Estimate parameters from the simulated data
data_summaries <- compute_data_summaries(clean_expr_data)
### Simulate hierarchical scRNA-seq data similar to input data
simulated_counts <- simulate_hierarchicell(data_summaries)
### Estimate power with a two-part hurdle mixed model
power_hierarchicell(data_summaries, n_genes = 100, n_per_group = 10)