bowtie terminate called error

[limericksean]

Hi,
I'm using bowtie 0.12.7 to align 100bp PE reads. I'm crashing with the error:

Too few quality values for read: -
are you sure this is a FASTQ-int file?
terminate called after throwing an instance of 'int'

Aside from finding out why this happened, I'd be particularly happy for bowtie to report this error as a warning and then continue with the run.
Is it possible to set a flag to get bowtie to throw the read instead of crashing?

Here's my call with arguments:
bowtie -X 20000 -q -p 12 -k 2 -m 2 --un no_align.fq --max mult_align.fq hg18.index -1 fastq_1 -2 fastq_2 align.aln

The last reported read before terminate is called can be found in no_align.fq (bowtie reports 2 files - 1 for each mate) :

tail -4 no_align_1.fq
@HWI-ST1096:2181WHPACXX:5:2316:16714:101026 1:N:0:TAAGGCG
NGATGAGATGATCGACGTCATCGGGGTGACCAAGGGCAAAGGCTACAAAGGGGTCACCAGTCGTTGGCACACCAAGAAGCTGCCCCGCAAGACCCACCGA
+
#1=?DDDDFFFFFDFII@@GFBCFEE@FDFFFEFFEFFFIIIIIIIIIEFFFDBDCCCCCBBB?BBBBBBBBBBBBBBBBABBBBBBBBBBBBBBBBB@9

&

tail -4 no_align_2.fq
@HWI-ST1096:2181WHPACXX:5:2316:16714:101026 2:N:0:TAAGGCG
CCAGAGCCCGCCGCTTCGTCTGCACCAGCAAGGACTTGCGGAGGGTGAGCGCCCGCTTCTTGGTTCCCACCGCACAGCCTTTCGGCATGACGGAGTCATT
+
1:?A1ADDF1@D1<CG<EHHEGHGJ@GD?)?DD2DG>=CHIIHFD;?=@@',33;@BBBA>>:@+>A@:A9&)09<(28CCDD(&&)08>A(&)0<9CCC


I've noticed other threads mentioning stretches of multiple B's at the end of sequences being a trigger for low quality reads. Not sure if this is the case here. Either way, can I force bowtie to not terminate on this instance if indeed this is the case.
One other thing, I have checked all reported align files (max-align, un-align & align) to see whether the next read in the original fastq file (i.e. after the one above) is actually reported just to make sure termination is in fact called & it is not there in any file. Therefore my run is incomplete.

Any ideas or fixes would be greatfully appreciated.

Thanks,
Sean.

[mastal]

bowtie terminate called error

Hi,

I don't think the problem is the read quality values, but rather the number of base qualities compared to the number of bases for that read.

I found a link to the part of the bowtie code that throws this error:

http://sauron.cs.umd.edu/bowtie/pat.cpp

Hope this helps,
Maria

[limericksean]

Thanks for the reply & tip.

However, this fills me with dread! I have to pre-check the qual & seq character lengths from the sequencing prior to alignments now. Man!
This has occurred in literally hundreds of different samples - it just can't be that they're all buggy.

S.