How to merge cufflinks result into a single matrix

[setia]

I am currently starting to run RNA-seq analysis using tophat and cufflinks. I am a beginner in this type of analysis. In my case I have RNA-seq reads (paired-end) from around 400 samples. I would be grateful if you can address the following questions:

1. As far as I know, Cuffmerge when we need to perform cuffdiff. In my study I want to classify the samples based on expression level obtained from cufflinks. I am wondering if Cuffmerge step is still needed if we are only interested in expression level ? (no differential expression analysis would be performed).

2. I thought that cufmerge will provide a sort of matrix containing of RPKM for all samples. But when I check the cufmerge result, I don’t see what I expected. I am wondering how we can combine/merge the FPKM for all samples into a single matrix as we usually have in gene expression using microarray.

3. I am still confused about the results of option -G and -g in cufflinks. If we specified -G reference.gtf, cufflink only calculate FPKM for only known transcript, but if we specified -g reference.gtf, it will include the reference transcripts as and novel genes and isoforms. Based on that I have the following questions:
a. If we specify -G, how can we distinguish which isoforms/genes are from the supplied reference annotation and which ones are novel genes/isoforms.
b. I checked the result from the defining -g and -G, the resulting values of FPKM for common isoforms are different. Is it expected?

[kopi-o]

1. I don't see why you should have to use cuffmerge.

2. You can probably do it with Cummerbund (http://compbio.mit.edu/cummeRbund/) - I haven't used it myself but rather prefer to combine the output files using a simple script, which could be written in any language but in my case is in R.

3. -g will have Cufflinks perform assembly (reference guided assembly) with the transcript annotation as a "guide", while -G will just do quantification against the GTF file and not do any assembly. In the -g mode, new transcripts will just have names like CUFF.123 while known transcripts will be tagged by the ID from the trasncript annotation. I don't remember the name(s) of the file(s) where you can see this, sorry.

[setia]

Hi Kopi-o, thanks for the response. Yes indeed we can write script in R to merge them. But I find out that for the -g mode number of known transcripts (names like ENS344) for a single sample is different as number of transcript when I used -G mode. I observed several new transcripts not listed in the result when I used -G mode where only used the known transcript. I also saw that some of known transcript from -G mode is not in the result using -g mode. Moreover the FPKM of some common transcripts are relatively different. It is confusing me.