keeping RNA-Seq reads mapping to repeats within genes

[Rachel Hillmer]

Hello friendly people,

I have a question about how to properly and efficiently deal with repeat regions within coding sequence when mapping RNA-Seq reads.

I have illumina 29bp reads (after trimming off a multiplex index), and am doing RNA-Seq with tissue from Arabidopsis thaliana, accession Col-0. I am currently using bowtie/tophat to perform my initial mapping.

To avoid signal from one gene leaking across to the signal from another, unrelated gene that happens to share one or more 29-mers (roughly, in actuality, 29 with as many as 2 mismatches), I throw away reads that map to multiple locations.
That is all fine and dandy, except that sometimes the repeats come from within the same gene. Transmembrane proteins, for example, are famous (in my book) for having long stretches of exact repeats. Throwing these away is undesirable, since throwing away clean signal always increases noise.

I can easily map my reads to all possible locations, and then write a perl script to sift through them afterwards, keeping multiple-mappers whose mapping locations are all within the same gene. (My organism is very, very well annotated.) But this seems like a basic function. Does anyone know of an option/switch in bowtie that allows what I am looking for (that I just can't find)?

Cheers!
~Rachel