R/ttestCtData.R

In HTqPCR: Automated analysis of high-throughput qPCR data

ttestCtData <-
function(q,
	groups	= NULL,
	calibrator,
	alternative	= "two.sided",
	paired	= FALSE,
	replicates	= TRUE,
	sort	= TRUE,
	stringent	= TRUE,
	p.adjust	= "BH",
	...)
{
	# Get the relevant data
	data <- exprs(q)
	# Collapse replicated genes if required
	if (replicates) {
		split.by <- rownames(data)
	} else {
		split.by <- featurePos(q)
	}
	data2 <- split(as.data.frame(data), split.by)
	feats <- split(featureNames(q), split.by)
	featPos <- split(featurePos(q), split.by)
	# Various checks
	if (length(groups) != ncol(data))
		stop("Dimensions of data and groups doesn't match\n")
	groups <- factor(groups, levels=unique(groups))
	if (length(levels(groups)) != 2)
		stop("Two factor levels required for \'groups\'\n")
	# Assign calibrator and target samples
	if (missing(calibrator))
		calibrator <- groups[1]
	g1 <- groups==calibrator
	g2 <- groups!=calibrator
	# Perform the t-test.
	t.tests <- lapply(data2, function(x) {
		x <- as.matrix(x)
		if (all(x==x[1,1])) {
			# Have to remove samples where all values are identical!
			list(p.value=1, statistic=NA, estimate=c(x[1,1], x[1,1]))
		} else {
			res <- t.test(x[,g1], x[,g2], alternative=alternative, paired=paired, ...)
			# Calculate mean for each group (not available if paired=TRUE)
			res[["estimate"]] <- c(mean(x[,g1]), mean(x[,g2]))
			res
		}})
	# Collect output
	means <- t(sapply(t.tests, "[[", "estimate"))
	colnames(means) <- c("meanCalibrator", "meanTarget")
	p.value <- sapply(t.tests, "[[", "p.value")
	t.value <- sapply(t.tests, "[[", "statistic")
	genes <- sapply(feats, "[[", 1)
	featurePos <- sapply(featPos, paste, collapse=";")
	# Make adjusted p-value
	adj.p.value <- p.adjust(p.value, method=p.adjust)
	# Fold change is calculated as ddCt, as well as 2^(-ddCT)
	cal <- grep(calibrator, colnames(means))
	FC <- means[, "meanTarget"] - means[, "meanCalibrator"]
	FC2 <- 2^(-FC)
	out <- data.frame(genes, featurePos, t.value, p.value, adj.p.value, FC, FC2, means, row.names=1:length(genes))
	# Indicate reliability of measure
	for (l in unique(groups)) {
		new.cat <- rep("OK", length(data2))
		old.cat <- split(featureCategory(q[,groups==l]), split.by)
		count.cat <- sapply(old.cat, function(x) sum(unlist(x) %in% c("Undetermined", "Unreliable")))
		cutoff <- ifelse(stringent, 1, ceiling(sum(groups==l)/2))
		new.cat[count.cat>=cutoff] <- "Undetermined"
		out[, paste("category", ifelse(l==calibrator, "Calibrator", "Target"), sep="")] <- new.cat
	}
	# Return output, sorted by p-value if requested
	names(out) <- c("genes", "feature.pos", "t.test", "p.value", "adj.p.value", "ddCt", "FC", colnames(means), grep("category", colnames(out), value=TRUE))
	if (sort)
		out <- out[order(out$p.value),]
	out
}