Simulate cell proportions using genotype principal components and population-level covariates

Function fits a multinomial regression model using the input genotype principal components (PCs) and other population-level covariates as training data and outputs simulated cell proportions.

Usage

simuCellProportion(
  sce,
  genoPC,
  new_genoPC,
  new_othercov,
  PCnum = 5L,
  cov_colnames = NULL,
  indiv_colname = "indiv",
  celltype_colname = "cell_type",
  cn_model_family = "lognormal",
  cn_meanlog = NULL,
  cn_sdlog = NULL,
  cp_model_family = "MN",
  cp_intercept = TRUE,
  ...
)

Arguments

sce

a SingleCellExperiment object.

genoPC

a data frame of individual by genotype principal components for sce input. The first column must be the variable for individual with same name as indiv_colname.

new_genoPC

a data frame of individual by genotype principal components for simulated individuals. The first column must be the variable for individual with same name as indiv_colname, followed by "PC1", "PC2", etc.

new_othercov

a data frame of the test data containing same additional covariates as in colData of sce.

PCnum

an integer scalar specifying the number of principal components used in multinomial regression.

cov_colnames

an optional string vector or scalar for the variable names to include in the cell proportion model. Variables must exist in both new_othercov and colData of sce.

indiv_colname

a string scalar to specify the variable in sce containing individuals.

celltype_colname

a string scalar to specify the variable in sce containing cell types.

cn_model_family

a string scalar to specify the model family used for total cell modeling. Currently only 'lognormal' from fitdistr is supported.

cn_meanlog

a numeric scalar for the mean parameter (on log scale) of the total cell number model. When cn_meanlog = NULL, the parameter is estimated from input data.

cn_sdlog

a numeric scalar for the standard deviation parameter (on log scale) of the total cell number model. When n_sdlog = NULL, the parameter is estimated from input data.

cp_model_family

a string scalar to specify the model family used for cell proportion modeling. Currently only 'MN' from dist is supported.

cp_intercept

a logical scalar for whether to include an intercept in the cell proportion model.

...

additional optional arguments.

Value

outputs a list with following elements:

simu_cov

a cell-by-covariate data frame of simulated cell types and corresponding individual.

cp_simu_df

a cell type-by-covariate data frame summarizing the simulate cell proportions, total cell numbers, and cells per cell types.

cp_modelfit

a fitted model object for the cell proportion model.

cn_modelfit

a fitted model object for the cell number model.

Examples

NULL
#> NULL