I am analyzing Tophat-Cufflink-Cuffdiff out put with CummeRbund package in R. I have read the CummeRbund manual and Tramper et al., (2012) papers about this program, however, I still have problems with some points:
> (scatterPlot_f_vs_m) summary(scatterPlot_f_vs_m)
but it seems it does not work, as the R value I am getting is always 1 and it sounds probably wrong to me. Another question regarding this is that we did not perform biological replicates (i.e., only one library per condition). so we are not sure whether this approach would be correct or not.
Just adding a few things to Thoma's answer:
2) If the normalization is correct you would expect the regression coefficient to be 1 (actually if it is different from 1 you should be worried). If the method Thomas suggest does not work just use the expression matrix from 1) and use the lm() function.
3) I completely agree with Thomas - no replicates => no statistical analysis (even though cufflinks try). The only option you have left is use a FC cutoff (but please dont term all with a FC > 2 for significant - since it is not), see below.
4) It might be that there are no genes termed significant by Cufflinks. Try to extract the data.frame with the differential analysis and filter on that:
diffAnalysis
Thanks a lot Tomas and Kristoffer for your comments and help.
Jose
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Bioinformatics
- Bioinformatic Tools