run.CSIDE.replicates: Runs CSIDE on a 'RCTD.replicates' object

In dmcable/RCTD: SpatialeXpressionR: Cell type identification and cell type-specific differential expression in spatial transcriptomics

Runs CSIDE on a RCTD.replicates object

Description

Identifies cell type specific differential expression (DE) as a function of the explanatory variable for each replicate. The design matrix contains an intercept column and a column of the explanatory variable. Uses maximum likelihood estimation to estimate DE and standard errors for each gene and each cell type. Selects genes with significant nonzero DE. Note: a minimum of three replicates are required for population mode.

Usage

run.CSIDE.replicates(
  RCTD.replicates,
  cell_types,
  explanatory.variable.replicates = NULL,
  X.replicates = NULL,
  cell_type_threshold = 125,
  gene_threshold = 5e-05,
  doublet_mode = T,
  weight_threshold = NULL,
  sigma_gene = T,
  PRECISION.THRESHOLD = 0.05,
  cell_types_present = NULL,
  fdr = 0.01,
  population_de = F,
  replicate_index = NULL,
  normalize_expr = F,
  test_genes_sig_individual = F,
  de_mode = "single",
  df = 15,
  barcodes = NULL,
  log_fc_thresh = 0.4,
  test_error = F,
  medv = 0.5,
  params_to_test = NULL,
  test_mode = "individual"
)

Arguments

RCTD.replicates	an RCTD.replicates object with annotated cell types e.g. from the run.RCTD.replicates function.
cell_types	the cell types used for CSIDE. Each cell type must occur at least 'cell_type_threshold', as aggregated by aggregate_cell_types
explanatory.variable.replicates	(only used for de_mode = single) a list of the named numeric vectors representing for each replicate the explanatory variable used for explaining differential expression in CSIDE. Names of the vectors are the SpatialRNA pixel names, and values should be standardized between 0 and 1.
X.replicates	(only used for de_mode = general) a list for each replicate of matrices containing the covariates for running CSIDE. The rownames represent pixel names and should be a subset of the pixels in the SpatialRNA object. The columns each represent a covariate for explaining differential expression and need to be linearly independent.
cell_type_threshold	(default 125) min occurence of number of cells for each cell type to be used, as aggregated by aggregate_cell_types
gene_threshold	(default 5e-5) minimum average normalized expression required for selecting genes
doublet_mode	(default TRUE) if TRUE, uses RCTD doublet mode weights. Otherwise, uses RCTD full mode weights
weight_threshold	(default NULL) the threshold of total normalized weights across all cell types in cell_types per pixel to be included in the model. Default 0.99 for doublet_mode or 0.95 for full_mode.
sigma_gene	(default TRUE) if TRUE, fits gene specific overdispersion parameter. If FALSE, overdispersion parameter is same across all genes.
PRECISION.THRESHOLD	(default 0.05) for checking for convergence, the maximum parameter change per algorithm step
cell_types_present	cell types (a superset of 'cell_types') to be considered as occuring often enough to consider for gene expression contamination during the step filtering out marker genes of other cell types.
fdr	(default 0.01) false discovery rate for hypothesis testing
population_de	(default FALSE) whether population-level DE should be run (can also be run later using the CSIDE.population.inference function.)
replicate_index	(default all replicates) integer list of replicate indices (subset of 1:N_replicates) to be run for CSIDE
normalize_expr	(default FALSE) if TRUE, constrains total gene expression to sum to 1 in each condition.
test_genes_sig_individual	(default FALSE) logical controlling whether on individual samples genes will be tested for significance.
de_mode	(default 'single', otherwise 'nonparam' or 'general') if 'single', calls run.CSIDE.single. If 'nonparam', calls run.CSIDE.nonparam. If 'general', calls run.CSIDE.
df	(default 15) for de_mode = nonparam, the degrees of freedom, or number of basis functions to be used in the model.
barcodes	for de_mode = nonparam, the barcodes, or pixel names, of the SpatialRNA object to be used when fitting the model.
log_fc_thresh	(default 0.4) the natural log fold change cutoff for differential expression
test_error	(default FALSE) if TRUE, first tests for error messages before running CSIDE. If set to TRUE, this can be used to quickly evaluate if CSIDE will run without error.
medv	(default 0.5) for single model, the cutoff value of explanatory.variable (after 0-1 normalization) for determining if enough pixels for each cell type have explanatory-variable greater than or less than this value (minimum cell_type_threshold/2 required).
params_to_test:	(default 2 for test_mode = 'individual', all parameters for test_mode = 'categorical'). An integer vector of parameter indices to test. For example c(1,4,5) would test only parameters corresponding to columns 1, 4, and 5 of the design matrix.

Value

an RCTD.replicates object containing the results of the CSIDE algorithm. See RCTD.replicates for documentation on the population_de_results, population_sig_gene_list, and population_sig_gene_df objects.

dmcable/RCTD documentation built on Feb. 24, 2024, 11:03 p.m.