Very large FPKM from cuffdiff that doesn't match read counts

fcchau · 02-09-2017, 11:02 AM

Hi Everyone,

I am relatively new to RNAseq, and hope can get some help from more experienced people here.

So I used cuffdiff to look at differential gene expression. I have 2 conditions and 3 replicates for each condition. When I look at the genes_read_group_tracking file, I found that for some genes, one of the replicates has very large FPKM values that doesn't match raw frags count.

For example this is what I see:

tracking_id condition replicate raw_frags internal_scaled_frags external_scaled_frags FPKM effective_length status
XLOC_009487 WT 0 4 4.21548 4.21548 1150.43 - OK
XLOC_009487 WT 1 5 5.39083 5.39083 1.14629 - OK
XLOC_009487 WT 2 8 7.56804 7.56804 1.60234 - OK
XLOC_009487 OE 1 2 2.29124 2.29124 0.476229 - OK
XLOC_009487 OE 0 5 4.84785 4.84785 0.995213 - OK
XLOC_009487 OE 2 5 4.07315 4.07315 0.77989 - OK

You can see that the FPKM for WT 0 is 1150 where as the raw frags is only 4.2. The other samples are fine. I observed this in multiple genes, and they don't always happen to the same sample. I also check the FPKMs from cufflinks, and they look normal. So seems that it's cuffdiff's problem.

Does anyone know why this happen and how to solve it? Appreciate the help!

fanli · 02-16-2017, 09:37 AM

I think there is a decent amount of evidence out there that cuffdiff is less than optimal for gene-level DE analysis.

See:
https://genomebiology.biomedcentral....-2013-14-9-r95
http://journals.plos.org/plosone/art...l.pone.0103207
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4914128/

If you are only doing DGE, I'd recommend edgeR or DESeq2

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Forward-reverse read counts don't match	Arvind Bhagwat	Core Facilities	1	04-18-2016 10:11 AM
Solution found: For Cuffdiff/links 2.0.2 Make cufflinks FPKM match Cuffdiff FPKM	NGSfan	RNA Sequencing	4	04-16-2013 08:10 AM
Varscan header doesn't match the number of columns in the data	rdeborja	Bioinformatics	3	01-30-2013 09:52 AM
Library QC- cntrl loci enrichment doesn't match pre and post library generation	yogioner	Sample Prep / Library Generation	0	12-03-2012 04:40 PM
Cufflinks convert FPKM to Read Counts	zee	Bioinformatics	0	03-08-2010 07:35 PM

02-09-2017, 11:02 AM	#1
fcchau Junior Member Location: Boston Join Date: Feb 2017 Posts: 1	Very large FPKM from cuffdiff that doesn't match read counts Hi Everyone, I am relatively new to RNAseq, and hope can get some help from more experienced people here. So I used cuffdiff to look at differential gene expression. I have 2 conditions and 3 replicates for each condition. When I look at the genes_read_group_tracking file, I found that for some genes, one of the replicates has very large FPKM values that doesn't match raw frags count. For example this is what I see: tracking_id condition replicate raw_frags internal_scaled_frags external_scaled_frags FPKM effective_length status XLOC_009487 WT 0 4 4.21548 4.21548 1150.43 - OK XLOC_009487 WT 1 5 5.39083 5.39083 1.14629 - OK XLOC_009487 WT 2 8 7.56804 7.56804 1.60234 - OK XLOC_009487 OE 1 2 2.29124 2.29124 0.476229 - OK XLOC_009487 OE 0 5 4.84785 4.84785 0.995213 - OK XLOC_009487 OE 2 5 4.07315 4.07315 0.77989 - OK You can see that the FPKM for WT 0 is 1150 where as the raw frags is only 4.2. The other samples are fine. I observed this in multiple genes, and they don't always happen to the same sample. I also check the FPKMs from cufflinks, and they look normal. So seems that it's cuffdiff's problem. Does anyone know why this happen and how to solve it? Appreciate the help!

02-16-2017, 09:37 AM	#2
fanli Senior Member Location: California Join Date: Jul 2014 Posts: 198	I think there is a decent amount of evidence out there that cuffdiff is less than optimal for gene-level DE analysis. See: https://genomebiology.biomedcentral....-2013-14-9-r95 http://journals.plos.org/plosone/art...l.pone.0103207 https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4914128/ If you are only doing DGE, I'd recommend edgeR or DESeq2