Search
Question: DESeq2 PCA plot: paired analysis
0
gravatar for g.atla
22 months ago by
g.atla0
g.atla0 wrote:

I am using DESeq2 to analysis rna-seq data with 8 biological replicates, which are paired samples. These samples are of primary cells, where variation between samples is expected. As this is a paired analysis, I am not removing batch effects.

 When I plot PCA, I could do not see that the samples are separated in to two groups.

Here is my code:

x <- read.table("filt_counts.txt", header=T, row.names=1)

subjects=factor(c(rep(1:8, each=2)))
treat <- as.factor(rep(c("High","Low"),8))

colData <- data.frame(colnames(x),subjects=subjects, treat=treat, row.names=1)
dds <- DESeqDataSetFromMatrix(countData = x, colData = colData, design = ~ subjects + treat)
design(dds) <- formula(~ subjects + treat)
dds <- DESeq(dds)

rld <- rlog(dds)
data <- plotPCA(rld, intgroup=c("treat", "subjects"), returnData=TRUE)
percentVar <- round(100 * attr(data, "percentVar"))

ggplot(data, aes(PC1, PC2, color=treat)) +
        geom_point(size=3) +
        xlab(paste0("PC1: ",percentVar[1],"% variance")) +
        ylab(paste0("PC2: ",percentVar[2],"% variance")) 

Should I trust the results despite having a PCA plot like above ? 
ADD COMMENTlink modified 22 months ago • written 22 months ago by g.atla0
1
gravatar for Michael Love
22 months ago by
Michael Love15k
United States
Michael Love15k wrote:

This just means that the subject effect is larger than the treatment effect. But you can still perform inference on the treatment effects using the ~subject + treat design. If you want, you can look at the results for significant genes using plotCounts, to see how treatment effects within subjects look.

ADD COMMENTlink written 22 months ago by Michael Love15k

Thanks Michael. 

ADD REPLYlink modified 22 months ago • written 22 months ago by g.atla0

What if I need to select few samples for further assays ? What would be the best approach ?

ADD REPLYlink written 22 months ago by g.atla0
I don't have a good answer for this. Remember, the observed data for samples and so their distances depends on underlying biology and also on technical factors like library preparation.
ADD REPLYlink written 22 months ago by Michael Love15k
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.2.0
Traffic: 123 users visited in the last hour