scMAGIC: A cell type classifier for single cell RNA sequencing data

In Drizzle-Zhang/scMAGIC: scMAGIC (Single Cell classification Model based on Auto-generated marker Genes and sImilarity Calculation)

A cell type classifier for single cell RNA sequencing data

Description

scMAGIC main function. User can provide reference data and query data, and scMAGIC can assign labels to cells in query dataset based on reference.

Usage

scMAGIC(exp_sc_mat, exp_ref_mat, exp_ref_label = NULL,
        single_round = F, identify_unassigned = T,
        atlas = c(NULL, 'MCA', 'HCL'), use.RUVseq = T,
        method_findmarker = c('COSG', 'Seurat'),
        percent_high_exp = 0.7, num_marker_gene = 100,
        cluster_num_pc = 50, cluster_resolution = 3, min_cell = 1,
        method1 = ('kendall', 'spearman', 'pearson', 'cosine', 'multinomial'),
        method2 = ('multinomial', 'kendall', 'spearman', 'pearson', 'cosine', 'randomforest'),
        corr_use_HVGene1 = 2000, corr_use_HVGene2 = 2000,
        threshold = 5, num_threads = 4, cluster_assign = F, simple.output = T)

Arguments

exp_sc_mat	The expression matrix of query data
exp_ref_mat	The expression matrix of reference data. If the type_ref is 'sc-counts', 'exp_ref_mat' is a single cell counts matrix where the row names correspond to gene symbols and the column names correspond to cell barcodes; if not, 'exp_ref_mat' is a matrix where the row names correspond to gene symbols and the column names correspond to cell type labels.
exp_ref_label	If 'type_ref' is 'sc-counts', 'exp_ref_label' is the vector including well-annotated cell labels corresponding to the cell barcodes of the 'exp_ref_mat'; if not, 'exp_ref_label' is NULL.
single_round	Whether to use single round annotation strategy, by default, scMAGIC use two-round annotation strategy.
identify_unassigned	Whether to label some cells as "Unassigned", by default, it is TRUE. If you think reference covers all cell types in query data, choose FALSE.
atlas	If reference is from mouse, it is 'MCA'; if reference is from human, it is 'HCL'; if no reference is avaiable, it is NULL.
use_RUVseq	Whether to use 'RUVSeq' to remove the batch effect between reference and atlas, by default, choose TRUE.
method_findmarker	Method to find marker genes.
percent_high_exp	In each cell type, the genes whose expression values are higher than "percent_high_exp" of genes are selected.
num_marker_gene	Number of each cell type's marker genes.
cluster_num_pc	Number of PCs used in clustering.
cluster_resolution	Resolution of the clustering algorithm, if you want get more clusters, you can give a larger resolution.
min_cell	If the number of validated cells with a reference cell type isn't less than "min_cell", these cells will be added into local reference.
method1	The methods of similarity calculation in first-round annotation, by default, choose 'kendall'.
method2	The methods of similarity calculation in second-round annotation, by default, choose 'multinomial'.
corr_use_HVGene1	Number of genes to select as top variable genes in similarity calculation of first-round annotation
corr_use_HVGene2	Number of genes to select as top variable genes in similarity calculation of second-round annotation
threshold	If confidence score is lower than 'threshold', we consider the corresponding label incorrect.
num_threads	Number of CPU used in calculation.
cluster_assign	Whether to annotate single cells by assigning a cell type to each cluster.
simple_output	Whether to output intermediate results.

Value

A dataframe including cell type labels and confidence scores.

Author(s)

Yu Zhang

Examples

##---- Should be DIRECTLY executable !! ----
##-- ==>  Define data, use random,
##--	or do  help(data=index)  for the standard data sets.

## The function is currently defined as
function (x)
{
  }

Drizzle-Zhang/scMAGIC documentation built on March 17, 2023, 2:31 a.m.