Closed rrydbirk closed 5 years ago
Also, if interested, I have scripts for Manhattan plot and Q-Q plot.
Hi Rasmus:
Sorry for the late reply. I was in a couple of deadlines (now finally submitted). >_<
I have checked the function, and yes I think it's a good design since if the second or third variables component's eigenvalues value is not hight, there is less worry about the batch effect on these components. One modification I want to do is join these figures side by side because many users run champ.SVD(), then directly check the result, so maybe it's a better idea to show them together. And maybe add the label of components on it.
I will modify above two issues, then push it to upstream Bioconductor repo, along with my greatest thank to you in the vignette.
I think the Manhattan plot and Q-Q plot would be useful as well. Actually recently one thing I am doing now is designing a comprehensive Quality Check tool for vast data sets in my group. (People keep arguing QC process in my group meeting ...). But since I did not use these two plots before, I better learn it first.
Again, thanks so much for your work. ^_^
Best Yuan Tian
Hi Yuan I'm glad that you found my suggestion useful. For Manhattan and Q-Q plots, I already have working code. Since they require DMP object to work (with adjPvalue = 1 to get all probes), I suggest incorporating them into champ.DMP function. I will create a new pull request for you to review. /Rasmus
Good Idea, I will merge it then.
Thanks Yuan Tian
Added scree plot (elbow plot) to champ.SVD() through old svdvis package (outdated). Added import ggplot2. Added new variable Splot and description in champ.SVD().