EstimateWeight: EstimateWeight function

In chichaumiau/DSA: Digital sorting of complex tissues for cell type specific gene expression profiles.

Description

Function to estimate the weight matrix. Based on the set of marker genes for each cell type, this function estimate the cell-specific proportions (weight) for each sample.

Usage

EstimateWeight(mix_ob, gene_list, method = "LM", l = 0, u = 2^34)

Arguments

data	data matrix of the mixture signals, with genes in row, cell type in column, in anti-log scale.
mix_ob	data matrix of the mixture signals, with genes in row, cell type in column, in anti-log scale.
gene_list	list of the length in the number of tissue types. Each list element contains gene symbols representing the tissue type.
method	methods used in estimating true signals for each tissue type. Default to 'LM' for linear regression. Other methods included 'QP_LM' (quadradic programming with constraint on the estimated parameter on linear regression)
l,u	values for the lower- (l) and upper- (u) bound used in setting the vector for values of b0 (bvec for solve.QP) in solving for quadratic programming. Defaults to 0 and 2^34 respectively.

Value

A list object containing two elements:

weight	matrix of estimated weight, with cell type in row, tissue types in column.
mse	means square error of the fitted linear model. mse is 'NULL' if method is 'QP_LM'.

See Also

GSM_QP, solve.QP

Examples

	## load package DSA
	library(DSA)
	
	## load sample data
	data('mix.signals')
	data('cell.gene')
	data('weight')
	
	# In this data, thr first three samples contain signal from only one cell	
	pure <- mix[, 1:3]
	mix <- mix[, 4:14]

	weight.orig <- weight
	weight <- weight[4:14, ]

	# Generate a list object set of marker genes for each cell type
	unique_cell_type <- unique(cell.gene[,2]) 
	gene_list <- list()
	for( i in 1 : length(unique_cell_type)){
		gene_list[[i]] <- cell.gene[cell.gene[,2] == unique_cell_type[i],1]
	} 
	names(gene_list) <- unique_cell_type

	# Estimate weight matrix
	estimated_weight <- EstimateWeight(2^mix, gene_list, method="LM")
	
	t.weight <- as.numeric(weight)/100
	e.weight <- as.numeric(t(estimated_weight$weight))
	limits <- range(t.weight, e.weight)

	plot(t.weight, e.weight, xlim=limits, pch=19, ylim=limits)
	cor(t.weight, e.weight)
	abline(a=0, b=1)

chichaumiau/DSA documentation built on Dec. 19, 2021, 3:56 p.m.