paired and unpaired analyses give same result

iDEP-SDSU / idep

Integrated Differential Expression and Pathway analysis

http://ge-lab.org/idep

126 stars 61 forks source link

paired and unpaired analyses give same result #60

Open tovahmarkowitz opened 5 years ago

tovahmarkowitz commented 5 years ago

I ran four comparisons in parallel on the same samples using limma-voom:

unpaired analysis on idep
paired analysis on idep
unpaired analysis using a personal script
paired analysis using a personal script The first three methods (all except my personal paired analysis) gave the exact same list of genes. Based upon PCA plotting, I was able to confirm that the "batches" were identified correctly but it isn't translating to the limma-voom analysis.

gexijin commented 5 years ago

Hi Tovah, Thank you for this detailed testing and report. We will look into it. Currently, we are re-writing iDEP, so that all steps can be reproduced with R Markdown. Please confirm that you are using limma-voom. And also the paired setting in iDEP is using the batch factor? Thanks again. Xijin (Xijin.ge@sdstate.edu)

tovahmarkowitz commented 5 years ago

Hi, There are issues with the paired/batch analysis for both versions of the current pipeline. The Rmd cannot incorporate the design table into the analyses at all. As for the Shiny version, when using limma-voom, the batch factor (for paired analyses) had no effect on the number of genes called. Tovah