Year 2019
Jump to navigation
Jump to search
TD Project
- Credit: Christopher Panlasigui
- Brief computational/statistical steps:
- Quantile normalization between all replicates (with log2 transformatuon)
- Linear model among the 3 groups
- Select top significant genes
- Map gene names
- Cluster and show by interactive heatmaps in R
# Quantile normalization
library(preprocessCore)
w2.mat <- as.matrix(w2[,3:17]) #convert original data to matrix
w2.mat.norm <- normalize.quantiles.robust(w2.mat, copy = TRUE, use.log2 = TRUE)
rownames(w2.mat.norm) <- w2$Geneid
# linear model
fits <- lmList(log.counts ~ group | geneID, data=w2.norm.melt)
lm.sum <- lapply(fits, function(x){
out<-summary(x);
fstat <- out$fstatistic;
pf(fstat[1], fstat[2], fstat[3], lower.tail=F)
})
p.df <- data.frame(gene=names(fits), p.val=as.numeric(lm.sum))
w2.out <- cbind(w2[,1:2], w2.mat.norm, p.df[,1:2])
# heatmap
heatmaply(td.mat, scale = "none" , cexRow = 0.50, colors = colorspace::diverge_hsv(16), branches_lwd = 0.3)
- Results