estimate_dbr_score: Estimate debris score per droplet
In marcalva/diem: Debris-Containing Droplet Identification using EM

estimate_dbr_score

R Documentation

Estimate debris score per droplet

Description

Create a debris score based on expression of debris-enriched genes. For each droplet, it sums the normalized read counts for these genes. Debris-enriched genes are calculated using a t-test between droplets in the debris and non-debris clusters. Only droplets in the test set are used. Genes with an adjusted p-value less than thresh_p, and a log fold change greater than thresh_logfc are included in the debris set. The number of genes included in the debris set can be capped by setting the max_genes parameter. This may help ensure that only the most specific genes are used in the case the many genes are significant in the DE test.

Usage

estimate_dbr_score(x, dbr_clust = c(1), thresh_genes = 200,
  thresh_p = 0.05, thresh_logfc = 0, max_genes = 5000,
  p_method = "fdr", sf = 1, verbose = TRUE)

Arguments

`x`	An SCE object.
`dbr_clust`	Specify additional debris clusters/
`thresh_genes`	Threshold for cluster mean number of genes detected. Clusters with an average number of genes detected less than this value are set to debris clusters.
`thresh_p`	P-value threshold for calculating differential expression between droplets in the debris and non-debris clusters.
`thresh_logfc`	Debris genes must have a log2 fold change greater than this value.
`max_genes`	The maximum number of debris genes to include, ranked by difference in proportion. This helps if a large number of genes are found significantly DE, and may improve the specificity of the debris score.
`p_method`	Method for p-value correction using `p.adjust`. expression between droplets in the debris and non-debris clusters.
`sf`	Scaling factor to multiple normalized counts by before log transformation. The normalized counts are used for the DE analysis.
`verbose`	Verbosity, indicating whether to show a progress bar.

Details

Differential expression is run by scaling the droplet counts to sum to sf and log transforming. Then, a t-test is run between droplets in the debris set and the cluster being tested.