MultiRD algorithm to deconvolute cell type proportions from a target bulk data

This is the main algorithm that MultiRD is relied on to implement deconvolution.

MultiRD.onegroup(
bulk.data,
list.marker,
celltype.unique,
subject.level.proportion,
population.level.proportion,
proportion.sd=1,
lambda.option=c(seq(from=0,to=0.075,length=15),10,50,100,500,1000),
tol=0.001,
iter.num=1000
)

Arguments

bulk.data

ExpressionSet object for a target bulk data
list.marker

A list of pre-specified marker genes corresponding to each cell type
celltype.unique

A list of cell types. It should match the order in list.marker
subject.level.proportion

A pre-specified cell type proportions for the target bulk data, which could be obtained from reference-based deconvolution approach.
population.level.proportion

A pre-specified population-level cell type proportions, which could be obtained from single-cell RNA-seq and external expression data from different studies, species, or data types
proportion.sd

an optional adjustment based on pre-specified standard deviation of cell-type proportion estimation. The default is 1 for each cell type.
lambda.option

a sequence of values for the tunning parameter
tol

a small constant used for convergence criteria. The default is 0.001
iter.num

The maximum number of iteration. The default is 1000.

Value

est.prop

a list containing estimated cell type proportions corresponding to each tuning value.

metrics

a sequence of goodness-of-fit values corresponding to each tuning value. The smaller the better.