mixtureModel: mixtureModel

In greenelab/miQC: Flexible, probabilistic metrics for quality control of scRNA-seq data

mixtureModel

Description

Function to fit a two-distribution mixture model on a SingleCellExperiment object.

Usage

mixtureModel(
sce,
model_type = "linear",
detected = "detected",
subsets_mito_percent = "subsets_mito_percent"
)

Arguments

sce	(SingleCellExperiment) Input data object.
model_type	(character) What type of model to generate. A linear mixture model ("linear") based on mitochondrial percentage and library complexity is recommended. B-spline ("spline") and two-degree polynomial ("polynomial") models are also supported. For a simpler model, a one-dimensional gaussian mixture model ("one_dimensional") based on mitochondrial percentage only is available. Default = "linear".
detected	(character) Column name in sce giving the number of unique genes detected per cell. This name is inherited by default from scater's addPerCellQC() function.
subsets_mito_percent	(character) Column name in sce giving the percent of reads mapping to mitochondrial genes. This name is inherited from scater's addPerCellQC() function, provided the subset "mito" with names of all mitochondrial genes is passed in. See examples for details.

Value

Returns a flexmix object with mixture model parameters, which is used to calculate posterior probability for each cell being compromised and make final filtering decisions.

Examples

library(scRNAseq)
library(scater)
sce <- ZeiselBrainData()
mt_genes <- grepl("^mt-",  rownames(sce))
feature_ctrls <- list(mito = rownames(sce)[mt_genes])
sce <- addPerCellQC(sce, subsets = feature_ctrls)
model <- mixtureModel(sce)

greenelab/miQC documentation built on Jan. 6, 2023, 12:16 a.m.