Its possible the sorting problem was a specific version and now gives
an error. The incorrect index caused by bad sequence lengths is a
recurrent problem, but I do not know what tool produces such headers.
Perhaps someone who has experienced this can chime in.

I'm not a samtools expert just sharing my experience on what has caused
this error int the past. It does seem that, as a general rule, that
these index problems result in errors from Picard (which the GATK uses),
while samtools can fail silently and sometimes and give you an unrelated
query region.

Jim

> Sending to galaxy-dev ...
>
> On Thu, Oct 27, 2011 at 5:51 AM, Jim Robinson
> <jrobinso@broadinstitute.org> wrote:
>> Hi Mike,
>>
>> Someone from the Galaxy team can perhaps give some insight on
>> what went wrong, I can comment on the error message from IGV.
>> That error is thrown from Picard, in every case I've investigated so
>> far it was traced to a problem with the index.
> Useful background re: "Error reading bam file. This usually indicates
> a problem with the index (bai) file. ArrayIndexOutofBoundsException:
> 4682 (4682)."
>
>> The most common causes are (1) a problem with the sequence
>> dictionary in the BAM header itself, specifically incorrect sequence
>> lengths,
> Any idea what tools produce that kind of thing?
>
>> and (2) indexing an un-sorted BAM. Apparently samtools will
>> make invalid indexes from such files without any complaints in
>> both cases. You can even use samtools tview on such files,
>> it happily will show you some random region when you query.
> That is news to me - I recall "samtools index" being recommended
> as a way to determine if a BAM files was sorted or not (error on
> unsorted, you get an index if it was sorted) and again from
> memory this is what Galaxy uses internally as part of preparing
> BAM files on upload.
>
> Might this be tied to a specific version of samtools? e.g. a
> possible regression?
>

>> I don't see a "Sort" step in your workflow, maybe that's the problem?
>>
>> Please CC me on any reply, I might miss it in the list.
>>
>> Jim
> Thanks,
>
> Peter