Understanding MA plot distortion
1
0
Entering edit mode
sp • 0
@sp-24958
Last seen 3.6 years ago
United States

Hello,

I have a question regarding the shape of my MA plot using deseq2. Basically I have two conditions that I am comparing (two types of blood cells within the mouse embryo), and three samples for each condition. I have noticed that in my dispersion, MA, and volcano plots there are a set of genes which are expressed in one condition but close to 0 in the other. I've included an MA plot after apeglm shrinkage, depicting that particular behavior. My suspicion is that it might be a closely related, contaminating/differentiating sub-population within the overall population that might be explaining these trends. I'm wondering whether any pre-filtering could be applied, or other techniques, that might be able to distentangle these genes from the overall analysis. I hope the explanation was ok, and thanks for any insight!

MA plot after shrinkage

DESeq2 MAplot deseq • 3.1k views
ADD COMMENT
0
Entering edit mode
@mikelove
Last seen 22 hours ago
United States

What if you find these genes by assessing:

cts <- counts(dds, normalized=TRUE)
idx <- res$log2FoldChange > X & rowMeans(cts[,dds$condition == "A"]) < Y

Here you would put in a value for X and Y, and for "A" that make sense for your experiment, perhaps by examining plotCounts of some example genes. And for filtering on LFC I would recommend to use the shrunken LFC from above.

ADD COMMENT
0
Entering edit mode

Great, thanks! I used an X val of >5 and Y val of <5 normed counts, and it seems like there's a strong differentiated lymphoid cell profile. A follow up question I have is: Is it justifiable to show these genes, and then somehow adjust for them/filter them out in downstream analyses? I Imagine single-cell analysis would've been preferable here but wondering if this can be salvaged. Thanks again!

ADD REPLY
0
Entering edit mode

I'm not sure best practice to filter them out, and don't have recommendations for adjusting -- these are just DE genes according to your sample groups. You can't really adjust if the composition is highly confounded with your groups.

You could just label the genes as those in functional categories (or using purified cell populations) related to lymphoid cells.

ADD REPLY

Login before adding your answer.

Traffic: 759 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6