What causes Linear relations on MAplots?

Question

[Gabriel]

Hello I am sort of a newbie in statistical science, so I hope this isn't a terrible question.

I am doing an analysis on example data, and I do the MAplot on my normalized counts from my DESeq2 results. This is what I get:

So we have a normal(i think) MAplot, of my two groups, each group has 3 samples.

But in bottom left corner, that is downregulated genes with low read counts we have a linear relation between read counts and Logfold change.

What could cause this?

Are these genes valid?

Another similar phenomena is seen when you compare two samples counts(before any rlog transform):

At the low read counts, again we have what looks like straight lines. Does it have something to do with the low sampling rate(counts) of these genes causing aliasing?

Best regards, Gabriel

[ATpoint]

How to add images to a Biostars post

Answer 1

[cmdcolin]

If you think in terms of algebra a line would be where y/x is a constant.

In the MA plot, Y=log(A)-log(B) and X=log(A)+log(B) so if (log(A)-log(B))/(log(A)+log(B))=constant then this is probably all the values from the gene set A that are 0 or gene set B are 0

[Gabriel]

Thank you. Your answer checks out! I used the identify() function to extract one of the gene names from the lines, and then I checked the raw data. It seems one group had all columns as 0, and the control group had all columns from 50-90 for that specific gene.

[ATpoint]

Please embed your images properly as suggested above ;-)

[Gabriel]

Sorry didn't see it until now, done.