program error on calling with hs38DH.fa

[gmagoon]

Just wanted to make note of an issue I encountered when calling with the entire hs38DH.fa (probably not a good idea, anyway). This is with trio calling and the --threads option. I suspect this may be related to the asterisk or colons in the HLA contig name.

[2018-12-11 23:56:22] <INFO>  chr19_GL949752v1_alt:987100            98.8%            1w 6d            3h 33m
[2018-12-12 00:07:43] <INFO>  chrUn_KN707874v1_decoy:2944            98.9%            1w 6d            3h 13m
[2018-12-12 00:13:29] <WARN> Skipping region chrUn_GL000214v1:117994-118601 as there are too many haplotypes
[2018-12-12 03:05:07] <INFO>      chrUn_KI270751v1:142912            99.0%            1w 6d            2h 55m
[2018-12-12 08:35:18] <INFO>               chr20:30136134            99.1%            1w 6d            2h 38m
[2018-12-13 04:14:52] <INFO>                            -             100%               2w                 -
[2018-12-13 04:23:25] <INFO> Removed 6734 temporary files
[2018-12-13 04:23:26] <EROR> A program error has occurred:
[2018-12-13 04:23:26] <EROR> 
[2018-12-13 04:23:26] <EROR>     Encountered an exception during calling 'failed to load contig
[2018-12-13 04:23:26] <EROR>     HLA-A*01:01:01:01'. This means there is a bug and your results are
[2018-12-13 04:23:26] <EROR>     untrustworthy.
[2018-12-13 04:23:26] <EROR> 
[2018-12-13 04:23:26] <EROR> To help resolve this error run in debug mode and send the log file to
[2018-12-13 04:23:26] <EROR> https://github.com/luntergroup/octopus/issues.
[2018-12-13 04:23:26] <INFO> ------------------------------------------------------------------------

[dancooke]

Thanks for reporting this Greg. The error is raised when Htslib does not successfully load a contig for VCF/BCF streaming. I'll investigate...

[gmagoon]

Thanks Dan. My open file limit was:

$ ulimit -Sn
16384

Not sure if that could be a factor here or not.

[dancooke]

Hi Greg, I'm having trouble reproducing this. Could you check if there are any reads mapped to HLA-A*01:01:01:01 in any of the BAMs please? Also, do you get the error if you just try to call this contig?

[dancooke]

Ignore that - have now reproduced. Will work on a fix!

[dancooke]

This was due to htslib not being able to parse :s in contig names (see here). Octopus creates temporary .bcf files - one for each contig - when using multithreading. During the merging of these temporaries after calling, the error was thrown. I've 'fixed' this by using ".vcf" temporary files for contig names with : (37a3329).

[gmagoon]

aha...great!...thanks very much Dan