8
$\begingroup$

I'm comparing the results that I obtain when doing a DE analysis with the Wald test and the likelihood-ratio test. One the thing that I've noticed is that there are many genes with 'beta' close to zero that are considered differentially expressed between the conditions.

I know that the likelihood-ratio test does't use the beta values to calculate the p-values, but I find it strange that transcripts with similar expression values between the conditions are being considered differentially expressed.

Volcano plot (Wald test q-values): enter image description here

Volcano plot (likelihood-ratio test q-values): enter image description here

Improve this question
$\endgroup$
5
  • 4
    $\begingroup$ I’ve never used Sleuth but there are several odd things about this data as far as I’m concerned: the y-axis on the Wald test plot for instance seems way off (inflated). The q-values in the LR plot are a lot more reasonable. $\endgroup$
    – Konrad Rudolph
    Aug 15, 2017 at 17:17
  • 1
    $\begingroup$ Cross-posted on biostars and the kallisto mailing list. $\endgroup$
    – Devon Ryan
    Aug 15, 2017 at 18:07
  • 1
    $\begingroup$ What does this look like if you plot Beta vs average TPM across all samples, excluding those with qval < 0.05? $\endgroup$
    – gringer ♦
    Aug 15, 2017 at 19:53
  • $\begingroup$ What is the actual question? Why are some genes with beta = 0 and qvalue < 0.05? $\endgroup$
    – llrs
    Sep 16, 2017 at 11:39
  • $\begingroup$ Yes, that's the thing I can't understand $\endgroup$
    – elsoja
    Sep 23, 2017 at 19:03

1 Answer 1

Reset to default
1
$\begingroup$

This question was answered here

Improve this answer
$\endgroup$
1
  • 2
    $\begingroup$ Could you post here also the solution to your problem? Just to avoid having to chase the anser from site to site $\endgroup$
    – llrs
    Sep 26, 2017 at 9:14

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.