DEseq2 differentially expressed genes have lower expression values than non-differentially expressed genes
2
0
Entering edit mode
lychen83 • 0
@lychen83-14436
Last seen 7.4 years ago

Hello everyone,

I used DESeq2 to do differentially gene expression analysis for 6 RNA-seq samples per species. 

tximport was used to input salmon mapping results to DESeq2.  I measured the expression value (txi.salmon$counts) for the differentially expressed genes per sample per species, and found that it is lower than that not suggested as differentially expressed genes.   For example, the average expression value for the differentially expressed genes is 100, while the value is 300 for the gene not suggested as differentially expressed.

I have tested 4 species, and found same problem. Could anybody explain or give suggestion?

Thank you

Chen Lingyun

 

deseq2 R • 1.6k views
ADD COMMENT
0
Entering edit mode
@mikelove
Last seen 2 days ago
United States

This is not a statistical problem. Counts of 100 is sufficient for detecting differential expression as long as the differences are relatively large and the biological variation is not so large.

This can be best visualized in an MA plot. What you are saying is that the DE genes are around normalized count of 100 on the x-axis. This can happen, if for example, one group has all 0's and the other group has counts ~200. Or even more subtle differences than on/off as long as the biological variation is low.

You can check plotCounts() for individual genes, in addition to the MA plot.

ADD COMMENT
0
Entering edit mode

Hi Michael,

Thanks for your great answer. It really helps me.

Is there any evidence/opinion shows that genes with overall low expression level across samples are more likely to be incorrectly identified as differentially expressed genes?  In other words, genes with high expression level are less likely to be identified as differentially expressed genes?

Thank you very much.

Best wishes,

Chen

ADD REPLY
0
Entering edit mode

"Is there any evidence/opinion shows that genes with overall low expression level across samples are more likely to be incorrectly identified as differentially expressed genes? " 

I don't think so. The methods take into account of the amount of precision from low and high counts genes. I'd recommend you look at plotCounts() for individual genes.

"In other words, genes with high expression level are less likely to be identified as differentially expressed genes?"

I think that the true DE genes can be skewed low, or highly expressed, or scatter uniformly. This is more about the biology.

ADD REPLY
0
Entering edit mode

Thank you so much. I will use plotCounts to see what happen

ADD REPLY

Login before adding your answer.

Traffic: 1001 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