This function is used to deconvolve bulk RNA-seq data using single-cell reference generated under a different clinical condition. Cell-type-specific differentially expressed (DE) genes are defined using a T statistics, calculated as the ratio of the mean to the standard deviation of the log fold change of cell-type-specific expression between conditions over all resampling iterations.

music2_prop_t_statistics(
  bulk.control.mtx,
  bulk.case.mtx,
  sc.sce,
  clusters,
  samples,
  select.ct,
  expr_low = 20,
  prop_r = 0.1,
  eps_c = 0.05,
  eps_r = 0.01,
  n_resample = 20,
  sample_prop = 0.5,
  cutoff_expr = 0.05,
  cutoff_fc = 2,
  cutoff_c = 0.05,
  cutoff_r = 0.01,
  maxiter = 200,
  markers = NULL,
  cell_size = NULL,
  ct.cov = FALSE,
  centered = FALSE,
  normalize = FALSE
)

Arguments

bulk.control.mtx

Matrix of expression for bulk data, control group

bulk.case.mtx

Matrix of expression for bulk data, case group

sc.sce

SingleCellExperiment for single cell data

clusters

character, the phenoData of single cell dataset used as clusters;

samples

character, the phenoData of single cell dataset used as samples;

select.ct

vector of cell types, default as NULL. If NULL, then use all cell types provided by the single cell dataset;

expr_low

numeric, cutoff on gene expression of the bulk data. Genes with mean expression across samples in bulk data < expr_low will be excluded from cell-type-specific DE gene detection. Default is 20;

prop_r

numeric, cutoff on cell type proportions for defining rare cell types. Cell types with mean proportion across samples in bulk data < prop_r will be characterized as rare cell types. Otherwise, will be characterized as common cell types. Default is 0.1;

eps_c

numeric, convergence cutoff for common cell types. The cell type proportion estimate is converged if absolute relative change of proportion estimates for the current iteration against the previous iteration < eps_c. Default is 0.05;

eps_r

numeric, convergence cutoff for rare cell types. The cell type proportion estimate is converged if absolute change of proportion estimates for the current iteration against the previous iteration < eps_r. Default is 0.01;

n_resample

numeric, number of resamples used for detecting cell-type-specific DE genes. Default is 20;

sample_prop

numeric, proportion of samples to be randomly sampled without replacement under each condition at each resampling iteration. Default is 0.5;

cutoff_expr

numeric, cutoff on gene expression over resamples. Genes with average cell-type-specific expression calculated over all resamples in the lower cutoff_expr quantile are excluded from cell-type-specific DE gene detection. Default is 0.05;

cutoff_fc

numeric, cutoff on fold change over resamples. Genes with absolute value of the mean fold change calculated over all resamples < cutoff_fc are excluded from cell-type-specific DE gene detection. Default is 2;

cutoff_c

numeric, cutoff on T statistics for defining cell-type-specific DE genes for common cell types. Genes with the value of T statistic in the upper cutoff_c quantile are considered as cell-type-specific DE genes. Default is 0.05;

cutoff_r

numeric, cutoff on T statistics for defining cell-type-specific DE genes for rare cell types. Genes with the value of T statistic in the upper cutoff_r quantile are considered as cell-type-specific DE genes. Default is 0.01;

maxiter

numeric, maximum number of iterations. Default is 200;

markers

vector or list of gene names. Default as NULL, i.e., use all genes that provided by both bulk and single cell datasets;

cell_size

data.frame of cell sizes. 1st column contains the names of cell types, 2nd column has the cell sizes per cell type. Default as NULL. If NULL, then estimate cell size from data;

ct.cov

logical. If TRUE, use the covariance across cell types;

centered

logic, subtract avg of Y and D;

normalize

logic, divide Y and D by their standard deviation;

Value

If MuSiC2 converges, return:

  • Est.prop: matrix, cell type proportion estimates.

  • convergence: logical, whether MuSiC2 converged or not.

  • n.iter: numeric, number of iterations.

  • DE.genes: vector, cell-type-specific DE genes being removed.

Or if MuSiC2 does not converge, return:

  • Est.prop: matrix, cell type proportion estimates.

  • convergence: logical, whether MuSiC2 converged or not.

  • id.not.converge: vector, sample ids that failed to converge.

See also