Genome Analysis Toolkit

Variant Discovery in High-Throughput Sequencing Data

GATK process banner

Need Help?

Search our documentation

Community Forum

Hi, How can we help?

Developed in the Data Sciences Platform at the Broad Institute, the toolkit offers a wide variety of tools with a primary focus on variant discovery and genotyping. Its powerful processing engine and high-performance computing features make it capable of taking on projects of any size. Learn more

HaplotyopeCaller SIGSEGV error in small region across multiple samples

0

4 comments

  • Avatar
    Gökalp Çelik

    Hi Michelle Noyes

    I see that you are using a JDK version above version 17 for 4.4.0.0 and above version 1.8 for 4.3.0.0. One recommendation we have (although you seem to have problems at a certain interval for only certain samples) is to use temurin JDK 17 from eclipse which is the officially supported version for 4.4.0.0. PairHMM Segfaults were noticed by our team and other users as well and most common solution is to use the temurin JDK (former adoptium) versions from https://adoptium.net/temurin website. 

    If this does not help we may ask you to send a slice of your bam files especially the problematic region so that we may test and see if we can reproduce the issue in our hands. 

    We hope this helps. 

     

    0
    Comment actions Permalink
  • Avatar
    Gökalp Çelik

    Hi Michelle Noyes

    As a current workaround for this problem, is it possible for you to try to call that region using the following parameter with HaplotypeCaller 

    --pair-hmm-implementation LOGLESS_CACHING

    This will disable OpenMP AVX accelerated pairHMM implementation and fall back to Java implementation which is slower but at least it may give us some idea. 

    Regards. 

    1
    Comment actions Permalink
  • Avatar
    Michelle Noyes

    Hello Gökalp Çelik 

    Adding the LOGLESS_CACHING parameter totally worked - it was able to march through the problematic region with no issues.

    I checked the output gvcf and it looked normal (the 100kb before the problem region had 2208 variant lines in the gvcf, compared to 2709 in the 100kb problem region).

    Thanks for the workaround! Do you have any sense of what the problem might be? 

     

     

    0
    Comment actions Permalink
  • Avatar
    Gökalp Çelik

    Hi Michelle Noyes

    We are glad that the workaround solved your problem. It is possible that there is an edge case in the native AVX accelerated library causing a segfault. We will relay this issue to GKL developers and let them check.

    Thank you for reporting back. 

    0
    Comment actions Permalink

Please sign in to leave a comment.

Powered by Zendesk