GSP_QP: GSM_QP function

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

Description

Function to estimate the true signals for each tissue type with quadratic programming.

Usage

GSM_QP(ob, weight, l = 0, u = 2^34, meq =0)

Arguments

ob	data matrix of the mixture signals (in anti-log scale), with genes in row, cell type in column.
weight	weight matrix, with cell type in column, tissue types in row.
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.
meq	default to zero (used to set meq for solve.QP)

Details

This functions depends on the solve.QP function from quadprog package. Its major job is to set up values for the needed parameters call solve.QP to obtain the solution.

Value

sol	estimated true signals for each gene (in row) in each cell type (in column).

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 <- weight[4:14, ]

	data <- as.matrix(2^mix)
	estimate_weight <- as.matrix(weight)
	estimate_weight <- estimate_weight/rowSums(estimate_weight)

	# Obtain deconvolution
	paraM <- GSM_QP(data, estimate_weight, l = min(data), u = max(data) , meq =0) 
	
	# check the estimated cell-type specific signals
	for(i in 1:3){
		print(cor(2^(pure[,i]),paraM[,i]))
	}

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