Standard error of log2FC from DESeq in time series experiment
2
0
Entering edit mode
Verena • 0
@verena-14992
Last seen 6.7 years ago

Dear all,


I am using DESeq2 to estimate fold changes of treatment versus control in a time series experiment (with biological replicates). I am a little puzzled about the standard deviations of the fold changes provided by DESeq2. These are almost constant over all timepoints, although it doesn't look like this when I judge the biological replicates in the output of plotCounts. Can someone please enlighten me how the standard errors of the log2FC computed? I'd be especially interested how the standard errors of a contrast c("Condition_treatment_vs_ctrl") vs a contrast like c("Condition_treatment_vs_ctrl", "Time24.Conditiontreatment") are computed (in a setting with design=~ Time + Condition + Condition:Time) (Here the 2 standard errors are almost the same, which I would not expect when looking at the data).

I hope I managed to ask the question in an understandable way. Your help is very much appreciated!

Thanks,
Verena

 

 

 

 

deseq2 log2fc standard error timecourse • 1.5k views
ADD COMMENT
2
Entering edit mode
@mikelove
Last seen 1 day ago
United States

The calculation of SE of beta for a contrast is in the Theory section of the vignette and in the DESeq2 paper.

Note that a single dispersion is calculated per gene. There is not a different dispersion parameter per group of samples.

ADD COMMENT
0
Entering edit mode

Thank you for very much for your quick answer.

Do I understand you correctly that with "group of samples" you mean replicates? Or do you want to say that the same dispersion is used per gene irrespective of the condition/timepoint? From the manual I understood that DESeq2 estimates the dispersion from replicates and therefore would have expected to obtain different dispersion values per timepoint.

Sorry, if I am slowly following only.

ADD REPLY
0
Entering edit mode

 One dispersion for the entire gene. This is alpha_i in the vignette and the paper. Note that this does not fix the variance to the same value for all counts, but more like a fixed coefficient of variation, so the variance for each count depends on its expected value.

ADD REPLY

Login before adding your answer.

Traffic: 812 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