DE fold change calculation issue
I have a edgeR results table linked to sample expected count values, but a question confused me a lot. Someone please just give me a simple explanation of how the number under A1, A2, A3, B1, B2 and B3 relate to the raw read numbers? For this given gene, the expect count may be (0,0,0 vs 5.7, 2.31,1.24) and the fold change is 1444.89x and the FDR is extreme small. Do I trust this? Thanks a lot!
Here is the example,
id Accession Version Fold change logFC FDR A1 A2 A3 B1 B2 B3
c78690_g3_i1 XP_007258149.1 1444.89 10.50 2.05E-16 0 0 0 5.7 2.31 1.24
• 3,630 views
•
link
updated
by
Ram
•
written
by
cecilin92 •
•
written
by
cecilin92 •
0 answers
No answers yet.
Log in to answer this question.
More posts like this
-
Create clustered bar chart in R
written by Prangan •
-
Calculate the Pearson correlation and associated p value for multiple variables
written by Cp.Recker •
-
Plot assigned / unassigned reads distance to 3'UTR
written by nlehmann
-
Calculating ratios by group in R
written by Sam •
-
Proteomics Data Analysis
written by osamashehzad.786 •
-
The best way to compare the clusters by different gene sets on the same RNA-Seq dataset?
written by huwenhuo •
-
How to combine multiple genes expression for survival analysis ?
written by gokce.ouz
-
DESeq2 - "all gene-wise dispersion estimates are within 2 orders of magnitude"
written by scchess
-
edgeR sample values
written by cecilin92 •
-
Potential Reason For Extreme Read Counts For A Single Replicate
written by alittleboy
Are you expecting a different answer than the one I gave you here?
No sir, I did appreciate your answer which help me a lot. Besides, I am wondering how the zero situation calculated but did not found in the tutorial. In SEQanswers someone gave a answer that adds a pseudocount of 0.125 to all observations, do you have any idea about this issue?
Yeah, I'd forgotten about that. It's the
prior.countoption.Thank you. Thanks to you the edgeR seems to be easier for me. Then I will dig out the calculation method behind prior.count.
One more question, based on such zero values, can I still trust in this differently expressed gene (for publishing)?
Thank you!
For this example I would be hesitant. The reason for this is that the counts in group B are still pretty low, so it's tough to gauge if this is really just noise. In any case, you should be validating some of the findings in independent samples with a different technology if the exact candidates are important (rather than if just changed pathways are important...in which case you'd then validate the pathway).