Implications from comparing cDNA to cds mapping

[sisch]

Hey there,

I've had an RNA Seq experiment on the 454 platform and during data analysis I came to wonder:

I mapped my reads (non-model organism) to a closely related genome using cDNA and coding sequences(cds). The fraction of mappable reads is 0.4 in cds and 0.6 in cDNA. So now I'm wondering whether this does infer that the 20% reads that mapped to cdna but not to cds are non-mRNA fragments (contamination of some sorts) or whether this implication is wrong

I feel like I'm overlooking something when I draw this conclusion but I can't point the finger to it.

Best,
Simon

[husamia]

what you mean by cDNA?

[Jeremy37]

I haven't dealt with RNA Seq data before... but isn't the difference just the non-coding portions of the mRNAs? I.e. 5' and 3' UTRs.

Mapping to coding sequences only would not include UTRs, whereas cDNA would.

[sisch]

It's just as Jeremy37 says. cDNA contains in addition to the coding parts also the UTRs.

Maybe there's just a huge bias towards UTRs when doing 454 sequencing.
For now i will be checking this phenomenon on different datasets.

[Eric Fournier]

It's just as Jeremy37 says. cDNA contains in addition to the coding parts also the UTRs.

Maybe there's just a huge bias towards UTRs when doing 454 sequencing.
For now i will be checking this phenomenon on different datasets.

A bias toward UTRs might arise from the protocol used to prepare your samples. For example, if you amplified your material using oligo-dT primers, you'll have a bias toward the 3'UTRs.

[Jeremy37]

Sisch, I still don't understand the problem. Why do you think it's wrong that a 0.4 fraction of reads maps to CDS and 0.6 maps to cDNA? You don't think the UTRs could be that long?

I don't know what the average is, but some UTRs are very long. You could calculate the expected fraction by comparing the same genes in CCDS vs. Refseq. To match your numbers you would expect the UTRs to comprise about half as much sequence as the CDS (since 0.6 is 50% more than 0.4).

[sisch]

@Eric: I just checked the library protocol again. It was a random primed amplification.
@Jeremy37: It's pretty much, that I didn't really understand, why I had a so much better mapping to cDNA than to cds. Now on second thought, the mapping to UTRs could make up for the 20%. Maybe I got a bit distracted, that less than 40% mapped, so I looked for a flaw in either sequencing or processing.

Thank you all for your contributions. I consider this thread as solved with the following conclusion for future readers:

A better mapping to cDNA in contrast to cds has no further implications on the sequencing, as long as the better mapping can be explained by sequence information derived from the UTRs (which won't map on cds obviously)