scMappR (version 1.0.11)

cwFoldChange_evaluate: Measure cell-type specificity of cell-weighted Fold-changes

Description

This function normalizes cwFold-changes by each gene to help visualize the cell-type specificity of DEGs. It then tests if a cell-type has a large change in correlation from bulk DEGs. Finally, it identifies genes that may be specific to each cell-type.

Usage

cwFoldChange_evaluate(
  cwFC,
  celltype_prop,
  DEG_list,
  gene_cutoff = NULL,
  sd_cutoff = 3
)

Value

List with the following elements:

gene_level_investigation

data frame of genes showing the Euclidian distances between cwFold-change and null vector as well as if cwFold-changes are distributed.

celltype_level_investigation

data frame of Spearman's and Pearson's correlation between bulk DEGs and cwFold-changes.

cwFoldchange_vs_bulk_rank_change

data frame of the change in rank of DEG between the bulk fold-change and cwFold-change.

cwFoldChange_normalized

cwFold-change normalized such that each gene sums to 1.

cwFoldchange_gene_assigned

List of cell-types where genes are designated to cell-type specific differential expression.

cwFoldchange_gene_flagged_FP

Mapped cwFoldchanges that are flagged as false-positives. These are genes that are driven by the reciprical ratio of cell-type proportions between case and control. These genes may be DE in a non-cell-type specific manner but are falsely assigned to cell-types with very large differences in proportion between condition.

Arguments

cwFC

A matrix or data frame of cell-weighted fold-changes of DEGs. Rows are DEGs and columns are cell-types.

celltype_prop

A matrix or data frame of cell-type proportions. Rows are different cell-types and columns are different samples. These cell-type proportions can come from any source (not just scMappR).

DEG_list

An object with the first column as gene symbols within the bulk dataset (doesn't have to be in signature matrix), second column is the adjusted p-value, and the third the log2FC path to a .tsv file containing this info is also acceptable.

gene_cutoff

Additional cut-off of normalized cwFold-change to see if a gene is cut-off.

sd_cutoff

Number of standard deviations or median absolute deviations to calculate outliers.

Details

cwFold-changes and re-normalized and re-processed to interrogate cell-type specificity at the level of the cell-type and at the level of the gene. At the level of the cell-type, cwFold-changes are correlated to bulk DEGs. The difference in rank between bulk DEGs and cwFold-changes are also compared. At the level of the gene, cwFold-changes are re-normalized so that each gene sums to 1. Normalization of their distributions are tested with a Shapiro test. Then, outlier cell-types for each gene are measured by testing for `sd_cutoff`'s mad or sd's greater than the median or mean depending on if the cwFold-change is non-normally or normally distributed respectively. Cell-types considered outliers are then further filtered so their normalized cwFold-changes are greater than the cell-type proportions of that gene and `gene_cutoff` if the user sets it.

Examples

data(PBMC_example)
bulk_DE_cors <- PBMC_example$bulk_DE_cors
bulk_normalized <- PBMC_example$bulk_normalized
odds_ratio_in <- PBMC_example$odds_ratio_in
case_grep <- "_female"
control_grep <- "_male"
max_proportion_change <- 10
print_plots <- FALSE
theSpecies <- "human"
toOut <- scMappR_and_pathway_analysis(count_file = bulk_normalized, 
                                      signature_matrix = odds_ratio_in, 
                                      DEG_list = bulk_DE_cors, case_grep = case_grep,
                                      control_grep = control_grep, rda_path = "", 
                                      max_proportion_change = 10, print_plots = TRUE, 
                                       plot_names = "tst1", theSpecies = "human", 
                                       output_directory = "tester",
                                       sig_matrix_size = 3000, up_and_downregulated = FALSE, 
                                       internet = FALSE)
                                       
cwFC1 <- toOut$cellWeighted_Foldchange
prop1 <- toOut$cellType_Proportions
DE <- bulk_DE_cors
eval_test <- cwFoldChange_evaluate(cwFC = cwFC1, celltype_prop = prop1,
                                    DEG_list = DE)