Hi,
so I would like to create an expression atlas. What I have is a mammalian genome and RNA-seq data (aligned BAM files) for 10 tissues. What I need are expression values (FPKM) for all genes I have annotated in there that are comparable across the 10 tissues.
My initial approach was to generate an 'augmented' annotation using cufflinks and cuffmerge and then cuffdiff to create the genes.fpkm_tracking file for FPKM values that I can 'compare' across samples. To clarify, I don't want to test for differential expression, the FPKM values should should just be on a comparable scale. I have done this twice now, finding things I would like to improve after each round. The problem is that cuffdiff takes forever to run this (3 weeks on 48 cores). So a third time is unlikely to happen.
Since I assume that most of the time used by cuffdiff is wasted on the DE analysis, are there any faster ways to get the data I am looking for? Would the individual cufflinks FPKM values be comparable, if I use the same annotation on all 10 samples? Cufflinks finishes in < 6-7h, so that would obviously be much nicer. However, I found a couple of threads where users observed deviations between cuffdiff and cufflinks FPKM values, so I am just wondering...
Regards,
Marc
so I would like to create an expression atlas. What I have is a mammalian genome and RNA-seq data (aligned BAM files) for 10 tissues. What I need are expression values (FPKM) for all genes I have annotated in there that are comparable across the 10 tissues.
My initial approach was to generate an 'augmented' annotation using cufflinks and cuffmerge and then cuffdiff to create the genes.fpkm_tracking file for FPKM values that I can 'compare' across samples. To clarify, I don't want to test for differential expression, the FPKM values should should just be on a comparable scale. I have done this twice now, finding things I would like to improve after each round. The problem is that cuffdiff takes forever to run this (3 weeks on 48 cores). So a third time is unlikely to happen.
Since I assume that most of the time used by cuffdiff is wasted on the DE analysis, are there any faster ways to get the data I am looking for? Would the individual cufflinks FPKM values be comparable, if I use the same annotation on all 10 samples? Cufflinks finishes in < 6-7h, so that would obviously be much nicer. However, I found a couple of threads where users observed deviations between cuffdiff and cufflinks FPKM values, so I am just wondering...
Regards,
Marc
