Hello,
We have performed an RNA-seq analysis with Tuxedo Tools. There are 2 conditions and 3 replicates for each, and we have detected differentially expressed genes with 0.01 q cut-off.
Now, it is time to get meaningful results by comparing with certain gene sets. I have used DAVID at the first step, then now, I am trying to use GSEA, but I have some questions in my mind.
- Should I load all the replicate FPKM values for each condition as input?
- Should I load all the genes without cut off?
Note: They say, you can use "Fpkm_trackingToGct" program from Genepattern. In this case you load the fpkm_tracking file of cuffdiff output and it gives you an output in .gct format which is going to be used as input in GSEA. In this case, that output includes the overall calculated value for 3 replicates (not all of them seperately). If this is the case, I assume I should use a cut-off when loading the genes because the program won't be able to calculate the variation btw replicates, but I also don't know if this leads to some bias.
We have performed an RNA-seq analysis with Tuxedo Tools. There are 2 conditions and 3 replicates for each, and we have detected differentially expressed genes with 0.01 q cut-off.
Now, it is time to get meaningful results by comparing with certain gene sets. I have used DAVID at the first step, then now, I am trying to use GSEA, but I have some questions in my mind.
- Should I load all the replicate FPKM values for each condition as input?
- Should I load all the genes without cut off?
Note: They say, you can use "Fpkm_trackingToGct" program from Genepattern. In this case you load the fpkm_tracking file of cuffdiff output and it gives you an output in .gct format which is going to be used as input in GSEA. In this case, that output includes the overall calculated value for 3 replicates (not all of them seperately). If this is the case, I assume I should use a cut-off when loading the genes because the program won't be able to calculate the variation btw replicates, but I also don't know if this leads to some bias.
Comment