Prediction pipeline bug(or no)!
Hi guys, I am opening this issue to solve the problem that I have with my pipeline. Unfortunately, I could not solve it, but you may have some good ideas. Problem: a few weeks ago we run my prediction pipeline on Naive CD4+ Tcell. The DE and diffTF analysis has been done by Sophia but without considering individuals(Donors). The result was ok.
DE_without_considering_individuals.pdf
Then we re-run DE and diffTF analysis but this time we consider individuals. The result got super worst and my prediction model only predicts positive values.
- I have checked the codes and couldn't find any bug.
- I have compared the DE data with each other and it looks not that bad.
- The correlation between these two is also around 90%.
- Here is also a comparison between p adjusted values (suggested by Wolfgang book)
my opinion: as I could not find any reason, I was thinking maybe just because of the R^2 when we are considering individuals are so low, the model predicts so bad which in our case everything is positive (which means nothing is wrong with DE analysis and prediction pipeline).
If anybody wants to take look at the codes, please let me know (I would appreciate it).