Combining FPKM values for a gene

john_nl · 02-13-2012, 06:29 AM

Hello,

Hopefully this is a simple question that gets a simple answer:

I just want to map my reads using bowtie to a well defined transcriptome in the form of a cDNA library. This works very nicely, but the problem is that after running it through cufflinks, it gives FPKM values for each transcript isoform that correspond to one gene. Essentially, I want one "score" for each gene- can i just sum FPKM values for each isoform for a given gene?

ta

adamdeluca · 02-13-2012, 06:48 AM

You can use the -G option to force cufflinks to only estimate expression of the reference transcripts you supply. Supply 1/gene.

kopi-o · 02-13-2012, 11:46 AM

You should be able to just sum the FPKMs.

Simon Anders · 02-13-2012, 12:10 PM

ARe you sure that cufflinks likes to work on data aligned to the [I]transcriptome[/I[? I suppose you mean by this that you supplied bowtie not with FASTA files for the genome but with files for the tranmscripts. This does not sound like the workflow it is designed for.

In any case, if you just want to know the expression strength of the genes, would it not be easier to count how many reads fall onto any exon of a given gene, rather than letting cuffdiff deconvolute this into transcripts, and you sum this up again.

natstreet · 02-15-2012, 09:39 PM

If you are mapping directly to a transcriptome I would maybe suggest trying RSEM and supplying it with a FASTA file of reference transcripts containing only one isoform per gene. The RSEM paper is here and the website here. RSEM is an R package that uses bowtie as a read mapper by default.

john_nl · 02-15-2012, 11:28 PM

Thank you very much for all of your answers. Much appreciated.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Different fpkm values for cuffdiff and cuffcompare	madsaan	Bioinformatics	3	12-12-2012 04:14 PM
Calculating p-values from FPKM?	Artur Jaroszewicz	Bioinformatics	16	10-25-2012 12:04 PM
Different FPKM values of cufflinks and cuffdiff	mrfox	Bioinformatics	5	10-17-2012 01:10 PM
Cufflinks and cuffdiff FPKM values	combiochem	Bioinformatics	12	10-13-2012 11:37 PM
FPKM minimum and maximum values ?	repinementer	Bioinformatics	2	11-18-2010 09:51 PM

02-13-2012, 06:29 AM	#1
john_nl Member Location: UK Join Date: Feb 2012 Posts: 13	Combining FPKM values for a gene Hello, Hopefully this is a simple question that gets a simple answer: I just want to map my reads using bowtie to a well defined transcriptome in the form of a cDNA library. This works very nicely, but the problem is that after running it through cufflinks, it gives FPKM values for each transcript isoform that correspond to one gene. Essentially, I want one "score" for each gene- can i just sum FPKM values for each isoform for a given gene? ta

02-13-2012, 06:48 AM	#2
adamdeluca Member Location: Iowa City, IA Join Date: Jul 2010 Posts: 95	You can use the -G option to force cufflinks to only estimate expression of the reference transcripts you supply. Supply 1/gene.

02-13-2012, 11:46 AM	#3
kopi-o Senior Member Location: Stockholm, Sweden Join Date: Feb 2008 Posts: 319	You should be able to just sum the FPKMs.

02-13-2012, 12:10 PM	#4
Simon Anders Senior Member Location: Heidelberg, Germany Join Date: Feb 2010 Posts: 994	ARe you sure that cufflinks likes to work on data aligned to the [I]transcriptome[/I[? I suppose you mean by this that you supplied bowtie not with FASTA files for the genome but with files for the tranmscripts. This does not sound like the workflow it is designed for. In any case, if you just want to know the expression strength of the genes, would it not be easier to count how many reads fall onto any exon of a given gene, rather than letting cuffdiff deconvolute this into transcripts, and you sum this up again.

02-15-2012, 09:39 PM	#5
natstreet Member Location: Sweden Join Date: Nov 2009 Posts: 83	If you are mapping directly to a transcriptome I would maybe suggest trying RSEM and supplying it with a FASTA file of reference transcripts containing only one isoform per gene. The RSEM paper is here and the website here. RSEM is an R package that uses bowtie as a read mapper by default.

02-15-2012, 11:28 PM	#6
john_nl Member Location: UK Join Date: Feb 2012 Posts: 13	Thank you very much for all of your answers. Much appreciated.