use bowtie for paired-end mRNA-seq reads?

[liux]

I am trying to map 10 million pairs of reads to the genome + splice junction templates. However Bowtie will not align most of the reads.

here is the command I used and the screen output (partial):

time bowtie -p 2 -v 2 -k 11 -m 10 -t --strata --best hg19.index -1 sample_10M_1.fastq -2 sample_10M_2.fastq sample.bowtie_aln.txt

# reads processed: 9999641
# reads with at least one reported alignment: 317813 (3.18%)
# reads that failed to align: 9679936 (96.80%)
# reads with alignments suppressed due to -m: 1892 (0.02%)

if I align them as single reads, majority reads are aligned:

bowtie -p 2 -v 2 -k 11 -m 10 -t --strata --best hg19.index -q sample_10M_1.fastq,sample$}_10M_2.fastq samplebowtie_aln_single.txt

# reads processed: 19999282
# reads with at least one reported alignment: 16697210 (83.49%)
# reads that failed to align: 3220837 (16.10%)
# reads with alignments suppressed due to -m: 81235 (0.41%)

I figured that it may be caused by the introns between the pair. what options I can use to make it work?

[Jon_Keats]

Why do you not want to run TopHat? Which is designed for this purpose largely using Bowtie on the front end. Alternatively, if you choose to do this for a specific reason you can try BWA instead of Bowtie, I've used it to do the same but you get lots or errors during the pairing step because of the introns. Alternatively, your strategy sounds like the one used in Alexa-Seq which works well.

[liux]

The intention is to compare how well different mRNAseq tools perform with a common dataset. ERANGE is one in the list, and ERANGE will use bowtie alignment. I could just use bowtie to align those as single reads and then use ERANGE.

[liux]

found a solution in the manual. since erange assumes two neighoring exons from one gene should be with in 20kb, the maximum 'insert' between a pair can be set by -X option, which is 20300 for this case. Now 72.9% reads are mapped.