detect_outlier: Detect outliers based on QC metrics
In scPipe: pipeline for single cell RNA-seq data analysis

Description Usage Arguments Details Value Examples

This algorithm will try to find comp number of components in quality control metrics using a Gaussian mixture model. Outlier detection is performed on the component with the most genes detected. The rest of the components will be considered poor quality cells. More cells will be classified low quality as you increase comp.

1 2	detect_outlier(sce, comp = 1, sel_col = NULL, type = c("low", "both", "high"), conf = c(0.9, 0.99), batch = FALSE)

`sce`	a `SingleCellExperiment` object containing QC metrics.
`comp`	the number of component used in GMM. Depending on the quality of the experiment.
`sel_col`	a vector of column names which indicate the columns to use for QC. By default it will be the statistics generated by 'calculate_QC_metrics()'
`type`	only looking at low quality cells ('low') or possible doublets ('high') or both ('both')
`conf`	confidence interval for linear regression at lower and upper tails.Usually, this is smaller for lower tail because we hope to pick out more low quality cells than doublets.
`batch`	whether to perform quality control separately for each batch. Default is FALSE. If set to TRUE then you should have a column called 'batch' in the 'colData(sce)'.

detect outlier using Mahalanobis distances

an updated SingleCellExperiment object with an 'outlier' column in colData

data("sc_sample_data")
data("sc_sample_qc")
sce = SingleCellExperiment(assays = list(counts = as.matrix(sc_sample_data)))
organism(sce) = "mmusculus_gene_ensembl"
gene_id_type(sce) = "ensembl_gene_id"
QC_metrics(sce) = sc_sample_qc
demultiplex_info(sce) = cell_barcode_matching
UMI_dup_info(sce) = UMI_duplication
# the sample qc data already run through function `calculate_QC_metrics`
# for a new sce please run `calculate_QC_metrics` before `detect_outlier`
sce = detect_outlier(sce)
table(QC_metrics(sce)$outliers)