MAplots after lfcshrink
1
0
Entering edit mode
mc • 0
@mc-23756
Last seen 4.8 years ago

Hello there,

I am having a problem when trying to generate MAplots after lfcshrink.

The following are my MAplots https://ibb.co/K9BZqpC https://ibb.co/ggyKr7v

When using lfcshrink, I see a strange pattern in the treatment samples at 30 min and 1hr compared to 2hrs (looking normal). I do expect that there are very few differences in the treatments 30 min and 1hr though. However, I am not seeing the trumpet shape in the MA plots when I generate using ggplot2 after lfcshrink. The plots without lfcshrink look okay, and when using plotMA function on lfcshrunk values, I see the trumpet shape but with y-axis in exponential log2fold changes.

I followed the standard DESeq2 analysis vignette, followed by lfcshrink, then to get the MA plots for each comparison. When I check the actual Log2FoldChange values in the results for each of the pairwise comparisons, I see that I am ending up with a large exponential fold changes in the treatments 30min and 1hr after lfcshrink, but the 2hrs timepoint looks fine after lfcshrink.

May I request you to suggest me if how could I fix this.

Thank you,

DESeq2 • 663 views
ADD COMMENT
0
Entering edit mode
@mikelove
Last seen 4 days ago
United States

I don't see anything strange here. You should set the ylim to be c(-2,2) across all to help see what is going on. Note that the exponentials are not large, they are very small.

ADD COMMENT
0
Entering edit mode

Thank you for that. Yeah they were very small exponentials. I did set the ylim to c(-2,2) https://ibb.co/K9BZqpC and thats where I am missing most of the genes for 30min and 1hr. The full range shows up only before lfcshrink, https://ibb.co/ggyKr7v. Could you help me understand the possible reason for such.

Thank you

ADD REPLY
1
Entering edit mode

This has been covered before on the support site, but it’s probably very difficult to find via search. apeglm only outputs LFC which are large if the evidence supports it. And it can be more conservative than the Wald test pvalue in terms of evidence from the counts.

ADD REPLY

Login before adding your answer.

Traffic: 852 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