Strange MA-plot using DESeq2

Question

[Cecelia]

I am running DESeq2 to perform a differential expression analysis two treatment conditions (Treat vs. Control) of a wasp infection in a beetle species. However I got some strange looking MA-plot when comparing some conditions.

There is an obvious set of outlier red dots in the down-regulated region that looks very strange. Can anybody explain what causes such results?

The count data was imported from Salmon output. The R command I used are below:

Importing data

filesevipus12h<- c("quant1.sf",    
 "quant2.sf",
 "quant3.sf",
 "quant4.sf",
 "quant5.sf",
 "quant6.sf")

names(filesevipus12h) <- c("Treat_12h_rep1",  "Treat_12h_rep2",  "Treat_12h_rep3",  "Control_12h_rep1",  
"Control_12h_rep2",  "Control_12h_rep3")
txifilesevipus12h <- tximport(filesevipus12h, type="salmon", tx2gene=tx2genepus)

DE analysis

ddsTCpus12 <- DESeqDataSetFromTximport(txifilesevipus12h, colData=colDatapus12,
                                   design=~ Treatment)
ddsTCpus12 <- DESeq(ddsTCpus12)

MA plot

pus12res <- results(ddsTCpus12, alpha=0.05)
plotMA(pus12res, main=paste0('G. pusilla'), ylim=c(-8,8))

Any help will be greatly appreciated. Thanks

I am having problem inserting images in biostar, the direct link to the DEseq2 image is below:

https://ibb.co/mgnzQ6

[Devon Ryan]

This looks mostly like the genes really do just change mostly in one direction. The normalization is a bit off center, but it's not terrible.

[cpad0112]

Please post image to understand the issue better.

[Devon Ryan]

I've corrected the link for the image.

Answer 1

[sHeyne]

The red cloud at the bottom looks like a contamination from some other cell/tissue type (in case of eg. mouse or human). In your case, all samples/replicates are from "whole body" beetles? Don't know how this then could happen...

In order to see if this effect relates to one or more replicates, you can check all pairwise correlation plots of your replicates. You should see a clear, reoccurring stripe next to the diagonal for certain samples. If you leave then out this sample(s), your MA-plot should look more normal.

[Cecelia]

Thanks for your suggestions! But could you explain more about the pairwise correlation plots of the replicates? What R function should I apply then?

[ATpoint]

Just out of interest, did you find out if the statement of sHeyne suggesting contamination was true? I do not really understand how one would make the connection from a cloud of apparently downregulated genes to contamination by other cells.

[Dunois]

"Whole body beetles" indicates that it wasn't just beetle, but beetle et al. that was/were sequenced. But sHeyne also claims there can't be any contamination (in this case) in the same breath. That answer sounds suspect.

Answer 2

[Renesh]

Easy to understand and visualize MA plot for gene expression data: https://reneshbedre.github.io/blog/ma.html