Modify marginal models

Modify the marginal model parameters of genes based on user inputs for cell-type-specific eQTLs.

Usage

modifyMarginalModels(
  marginal_list,
  eqtlgeno_list,
  features,
  celltype,
  neg_ctrl = FALSE,
  mean_log2fc = 0,
  eqtl_log2fc = mean_log2fc,
  eqtl_reverse = FALSE,
  mean_baseline = NULL,
  eqtl_baseline = NULL,
  mean_baseline_only = FALSE,
  eqtl_baseline_only = FALSE,
  disp_scaling = FALSE,
  celltype_colname = "cell_type",
  snp_colname = "snp_id",
  verbose = TRUE,
  debug = FALSE,
  ...
)

Arguments

marginal_list

A list of marginal model objects.

eqtlgeno_list

A list of eqtl genotypes.

features

A scalar or vector of features (ie. genes) to apply the modifications.

celltype

A string to specify the cell type in which to make the modification.

neg_ctrl

A logical value for whether to set a negative control eQTL (ie. a non-eGene). This option sets the conditional means to be identical across genotypes (0, 1, 2). If neg_ctrl = TRUE, the mean_log2fc option will still be applied if set, but eqtl_log2fc will be overidden and have no impact. Default is FALSE.

mean_log2fc

A numeric scalar or vector for the log2 fold-change parameter to increase or decrease the conditional mean at genotype 1 \(\mu_{1}\) in a cell type. Default is mean_log2fc = 0 (no parameters are modified and uses estimated parameters from the fitted marginal model).

eqtl_log2fc

A numeric scalar or vector for the log2 fold-change parameter to increase or decrease the slope of eQTL effect in a celltype. The eQTL slope is defined as the difference between the conditional mean at genotype 1 and genotype 0 (\(\mu_{1}\) - \(\mu_{0}\)). Default is eqtl_log2fc = mean_log2fc (eQTL slope is scaled the same as the conditional mean log2 fold-change).

eqtl_reverse

A logical value to determine whether the eQTL slope trends in the reverse direction (TRUE) or same (FALSE). Default is FALSE.

mean_baseline

A numeric scalar or vector to specify the minimum conditional mean at genotype 1 \(\mu_{1}\). If mean_baseline_only = FALSE, then the conditional mean will be the maximum of the fitted (estimated from marginal model) and the mean_baseline value. Otherwise, the conditional mean will be set to the mean_baseline value. Default value is NULL.

eqtl_baseline

A numeric scalar or vector to specify the minimum eQTL slope between genotype 1 and 0 (\(\mu_{1}\) - \(\mu_{0}\)). If eqtl_baseline_only = FALSE, then the eQTL slope will be the maximum of the slope of fitted (estimated from marginal model) and the eqtl_baseline value. Otherwise, the eQTL slope will be set to the eqtl_baseline value. Default value is NULL.

mean_baseline_only

A logical scalar or vector to force the conditional mean (in linear prediction) at genotype 1 \(\mu_{1}\). Default is FALSE.

eqtl_baseline_only

A logical scalar or vector to force the eQTL slope between genotype 1 and 0 (\(\mu_{1}\) - \(\mu_{0}\)). Default is FALSE.

disp_scaling

A string or logical value to specify the dispersion-mean scaling for certain parametric models. Current options are either "linear", "quadratic", or NULL. If FALSE, the dispersion parameter is not changed. The default is FALSE. (NOTE: currently only applicable to the negative binomial model.)

celltype_colname

A string for cell type variable name.

snp_colname

A string for SNP id variable name.

verbose

A logical value for whether to output messages related to modified parameters. Default is TRUE.

debug

A logical for whether to output a mod_list list in addition to marginal_list.

...

Additional options used. Currently supported:

mod_scale

A string to specify on what scale the modified response will be on. The two options are 'response' or 'link'. The default is 'response'.

Value

A list of marginal models similar to marginal_list input. When debug = TRUE, the output have a mod_list list containing intermediate objects in addition to marginal_list.

Examples

NULL
#> NULL