Hi,

I am trying to understand PCA plot I generated using Diffbind.

I used this code

ss <- read.csv("samplesheet.csv")
obj <- dba(sampleSheet=ss,  scoreCol=5, minOverlap=1)


# get counts (mapQCth=0 was kept to match with results from DESeq2/featurecounts, summits=FALSE for counting over full peak )
obj <- dba.count(obj, minOverlap=1, score=DBA_SCORE_READS, summits=FALSE, mapQCth=0)
obj <- dba.normalize(obj)
obj <- dba.contrast(obj, minMembers = 2,categories=DBA_CONDITION)
obj <- dba.analyze(obj)
dba.plotPCA(obj, attributes=DBA_CONDITION, label=DBA_REPLICATE, score=DBA_SCORE_NORMALIZED)

It seems this PCA plot generated by dba.plotPCA is not accounting for normalized reads rather it is using raw counts to plot PCA. I rectified it separate analysis with quantification with featurecounts/DESeq2 and plotting PCA on both vst transformed counts and raw counts. The PCA obtained from diffbind looks more similar to raw counts than vst.

So what I am doing wrong in this code,

dba.plotPCA(obj, attributes=DBA_CONDITION, label=DBA_REPLICATE, score=DBA_SCORE_NORMALIZED)

Am I missing something? Will it also influence differential analysis?

Please help.

Thank you