selectTargetGenes: Select target genes

In argschwind/TAPseq: Targeted scRNA-seq primer design for TAP-seq

Select target genes

Description

Select target genes that serve as markers for cell populations using a linear model with lasso regularization. How well a selected set of target genes discriminates between cell populations can be assessed in an intuitive way using UMAP visualization.

Usage

selectTargetGenes(object, targets = NULL, expr_percentile = c(0.6, 0.99))

plotTargetGenes(object, target_genes, npcs = 15)

Arguments

object	Seurat object containing single-cell RNA-seq data from which best marker genes for different cell populations should be learned. Needs to contain population identities for all cell.
targets	Desired number of target genes. Approximately this many target genes will be returned. If set to NULL, the optimal number of target genes will be estimated using a cross-valdation approach. Warning: The number of target genes might end up being very large!
expr_percentile	Expression percentiles that candidate target genes need to fall into. Default is 60% to 99%, which excludes bottom 60% and top 1% expressed genes from markers.
target_genes	(character) Target gene names.
npcs	(integer) Number of principal components to use for UMAP.

Value

A character vector containing selected target gene identifiers.

Examples

library(Seurat)

# example of mouse bone marrow 10x gene expression data
data("bone_marrow_genex")

# identify approximately 100 target genes that can be used to identify cell populations
target_genes <- selectTargetGenes(bone_marrow_genex, targets = 100)

# automatically identify the number of target genes to best identify cell populations using
# cross-validation. caution: this can lead to very large target gene panels!
target_genes_cv <- selectTargetGenes(bone_marrow_genex)

# create UMAP plots to compare cell type identification based on full dataset and selected 100
# target genes
plotTargetGenes(bone_marrow_genex, target_genes = target_genes)

argschwind/TAPseq documentation built on Feb. 9, 2024, 8:20 p.m.