UTR's in Bowtie2 Global Alignment?

[Coryza]

I've got paired-end MiSeq RNA data from Illumina, with as far as I know adapters/primers/barcodes clipped off. When mapping this data with bowtie2 against the human reference mRNA's I don't find even 1 global alignment in +- 500 sequence pairs.

Could this be due to the UTR's probably still in the sequences and not in the references? Or doesn't that matter for global alignment?

[dpryan]

1) You probably don't want to do that. Aligning directly against the ref mRNA will probably just make life more complicated downstream.

2) Try aligning the pairs as singletons (i.e., unpaired) and see what happens. Sometimes read #2 in a sample is complete crap and that keeps things from aligning. Also, perhaps the insert size is bigger than the default (500) or the edit distance is larger than allowed by default (i.e., tweak --score-min). UTRs should still be in the reference, it wouldn't make any sense to remove them. If they're gone, then yes, that'll seriously decrease the alignment rate, particularly in end-to-end (i.e., global) alignment.

[Coryza]

Why is it a bad idea to align against the ref mRNA? What should I align against too then?

[dpryan]

The genome. Alternatively, you could use tophat2, which will align against both in turn and only return alignments in genomic coordinates.

[Coryza]

Yea after bowtie2 I was planning to use tophat2. But I still don't understand why its a bad idea to don't map our mRNA data against the mRNA references.

[dpryan]

It just makes your life more complicated. Firstly, many of the transcripts will be overlapping, so when you want to determine how unique an alignment actually is then you need to recalculate everything. Secondly, the returned coordinates will be in transcript-space, which is rarely useful for anything.