Merge three assembled files into one.

[netpumber]

Hello.

I'm stack at the procedure for analyze some RNA-seq data. The experiment is the following.

1)
Plants have been transformed for silence one specific gene. From that transformation 3 plants (different events) occurred and checked that the gene has been truly silenced.

Three wt (Wild type) plants also used for the later comparison.

2)
These 6 plants have been RNA sequenced, in order to calculate differential expression.

3)
The workflow i follow is next:

a) Alignment of the reads, using the reference genome.
As output i got 6 .bam files.

b) Assembly of the reads, using again the reference genome.
As output i got 6 files.

Here is where i stack and i want your opinion.
Because three files of them (which are from the transformed plants) are from different plants , i suppose that they will not have exactly the same results. For example, a specific plant will lack a gene that the other 2 plants will have it.

The same issue will be appear in Wild type plants too.

So because i don't want (i don't think that it's right) to compare for differential expression ,
WT plant 1 with Transf. plant 1
WT plant 2 with Transf. plant 2
WT lpant 3 with Transf. plant 3

I'm asking if its possible to merge the three results of transformed plants into one file and the three results of WT plants into one file and then compare those two new files.

Thank you.

[Brian Bushnell]

That sounds like a very bad idea to me. Each species should be compared only with another plant of the same species, or you'll end up with data that has basically no meaning. Say, for example, species 1 and 2 had expression of the gene doubled and species 3 had expression totally shut off; your results might indicate that there was absolutely zero change, which would be wrong.

[netpumber]

I don't understand why you think that plants aren't of the same species. They are all tomatoes.

Anyway. I'll try it to make it more clear. Let's say that

in 1st transformed tomato plant, geneX has been expressed 2 times
in 2nd transformed tomato plant, geneX and geneY have been expressed 1 time
in 3rd transformed tomato plant, geneP has been expressed 3 times

what i thought is to create a new file containing the mean of expression e.g:

geneX has been expressed (2+1+0)/3 = 1 time
geneY has been expressed (0+1+0)/3 = 0.33 time
geneP has been expressed (0+0+3)/3 = 1 time

Then to do the same with the wild types and compare those new files.
You still think that its a bad idea ?

Finally, your thought is to compare each plant individually such as ? :
WT plant 1 with Transformed plant 1
WT plant 2 with Transformed plant 2
WT plpant 3 with Transformed plant 3

Thank you.

[dpryan]

There are two general ways in which this is done:

A. Each sample is aligned and assembled followed by a merge step (this is what the cufflinks pipeline does). Quantification is then performed on the merged alignment.
B. The mapping results are merged together, the merged results are assembled, and each sample is then quantified accordingly.

Option B can be used for any assembler, since there aren't many that allow merging a la cufflinks (cufflinks has had a variety of problems, which is why many of us avoid it).

For the actual comparisons, don't use fpkms or try to make up your own metric (as outlined in your reply to Brian). There are a variety of issues that need to be dealt with and you're much better off using DESeq2, edgeR, limma/voom, etc. than making your own method.

BTW, you do not want to compare individual plants. That has no biological meaning (who cares if two plants differ, that's not predictive of anything).

[netpumber]

Actually, i aligned reads with tophat2 and assembled them with stringtie (instead of cufflinks). As you said, this pipeline is followed by a merge step. Did you mean the cuffmerge ? If you mean so, are you sure that cuffmerge is able to get as input more than two files (in our problem 12 ) and merge them ?

[dpryan]

Yes, I mean cuffmerge from the cufflinks suite. Yes, it can accept an arbitrary number of files, so just give it all of them.

[netpumber]

Thank you very much.
I'll give it a try.