I want to identify highly expressed genes in human corneal (normal) cell types. For that, I retrieved transcriptome profiles of different studies from GEO datasets and carried out till CPM normalization. In the PCA plot, there is inter-study sample variability. How to reduce or ignore this effect?
Hi Jeyaprakash Kumar
the difficulty is to take data from database from different studies with obviously different qualities, bias and results. one way would be to choose more studies and delete the most different ones, or to use in your code a function to reduce those batch effects and increase PCs, as SVA:
https://bioconductor.org/packages/devel/bioc/vignettes/sva/inst/doc/sva.pdf
all the best
fred
Thank you Frederic Lepretre . I will definitely try it out and get back to you.
Hi Dipanka Tanu Sarmah and Frederic Lepretre
I did batch effect correction using R (ComBat function). PCA plot,after batch effect correction, is attached. Please share your comments on this plot, which will be helpful for me.
Thanks
Jeyaprakash
Hi Yat Sen Wong
I did ComBat in R. I will try Multibac and get back to you. Thank you.
Hi
maybe you could filter some samples (as S20, GSM1020215 and his neighbor).
fred
- Badges
- Science topic