Gene prediction from RNA-Seq assembly

[bob-loblaw]

Hi all,

I've seen in a few different places that gene prediction is included as part of an RNA-Seq assembly analysis framework. But, in theory at least, shouldn't the contigs correspond to coding regions? (assuming that a poly-A capture or rRNA depletion was used).

If a large portion of the reads are aligning to regions assembled that aren't predicted to be coding indicate DNA contamination?

Thanks

[bastianwur]

If the data is prokaryotic: No, you have the polycistronic mRNAs, so a transcript will easily be on average 2 genes.
You also have the UTRs, and with small intergenic distance you'll then assemble multiple independent transcripts into one contig.
You might also have overlapping genes on both strands.

I'd worry about DNA contamination when the contigs really get too long.
In one of our recent dataset we had 3 samples which were apparently DNA, and we contigs of 500kb -> that's definitely wrong.
With only a few genes on a transcript, I'd not worry.