clarification of the way of converting standard error to confidence interval for logFC in DESeq2 result table
1
0
Entering edit mode
Mike • -60
@1df4ed7f
Last seen 7 days ago
United States

Enter the body of text here Hi, there:

I read the previous posts about deriving confidence interval for fold change using DESeq result at URLs: 95% Confidence intervals for Fold Changes computed by DESeq2 and Error bars DESeq or DESeq2 fold change

one comment from Micheal Love mentioned: Estimated standard errors for the estimated coefficients on the log2 scale are given by the lfcSE column. Yes, you can convert these to fold change errors using this formula: 2^(coef - SE) and 2^(coef + SE).

If you multiple lfcSE by normal quantiles (e.g. qnorm(.025)), you can construct confidence intervals for the coefficients. ....

my concerning point that needs a bit clarification or confirmation is whether this referred to logFC or FC:

one of my own result table of an interested contrast from DESeq2 is shown below as an example:

SYMBOL baseMean log2FoldChange lfcSE pvalue padj

Igfbp3 6947.54841 7.586917789 0.323830741 2.08E-123 3.81E-119

Gap43 2411.144079 4.597075539 0.195529679 3.43E-123 3.81E-119

Bmp2 1134.843753 7.659549924 0.351744204 3.54E-107 2.62E-103

.......

for gene Igfbp3, the log2FoldChange is 7.586917789, lfcSE is 0.323830741, so to derive 95% confidence internal (95%C.I.) accordinyl based on formula for 95%C.I., we can do following: [7.586917789-0.323830741qnorm(0.975), 7.586917789+0.323830741qnorm(0.975)] = [6.952221, 8.221614].

this calcualtion is based on logFC and and I read from the manual: The lfcSE gives the standard error of the log2FoldChange. and so the derived 95%C.I shall be for logFC, is my understanding correct? and for FC, shall be simple as 2^(6.952221) and 2^(8.221614)? just want to double check to make sure. also another question is: whether I used Log fold change shrinkage or not would impact the calculation or not? say if I do either res <- results(dds, contrast=Con1); or resLFC <- lfcShrink(dds, contrast = Con1, type="ashr"), the way of deriving 95% C.I for logFC would be different, or any concerns or issue?

Thanks so much for your help!

Best

Mike

Code should be placed in three backticks as shown below


# include your problematic code here with any corresponding output 
# please also include the results of running the following in an R session 

sessionInfo( )
DESeq2 Confidence.Interval • 120 views
ADD COMMENT
0
Entering edit mode
@mikelove
Last seen 8 hours ago
United States

Yes, LFC + Q * lfcSE gives the CI for the LFC, and then you raise 2^[LFC lower, LFC upper] to get the CI for the FC.

whether I used Log fold change shrinkage or not would impact the calculation or not?

This is up to you. If you use the shrinkage, then it is not a confidence interval, but a credible interval.

ADD COMMENT
0
Entering edit mode

Really Appreciated your input and confirmation, Michael! very helpful! Mike

ADD REPLY

Login before adding your answer.

Traffic: 313 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