GATK4.0.9.0 undetected a delins site

Answers

• bshifaw moonMember, Broadie, Moderator admin

  May 29 edited June 2

  Hi @qiuqiu87,

  "HaplotypeCaller performs a local reassembly and realignment of the reads in the region surrounding potential variant sites" then the tool calls the variants. If you would like to know more about how haplotypecaller calls variants you can read through Local re-assembly and haplotype determination.
  The site in question may not have met the default threshold to be considered a variant by Haplotypecaller due to quality of evidence. However, the tool does have parameters that can be tweaked to be more sensitive to calling variants which can be found in the tool documentation page (warning, this increases the chance of false positives).

  If you believe the tool is misbehaving please read the following doc Generate a "bamout file" showing how HaplotypeCaller has remapped sequence reads to generate a bamout and a IGV screenshot of the region not being called and post it here.

  Post edited by bshifaw on June 2

  0 · Share on Twitter Share on Google+
• qiuqiu87 Member ❠

  May 31

  Hi @bshifaw
  The screenshot showed the raw bam and bamout about the not being called region. The two bam files have reads support the variant.

  image.png 46.6K

  0 · Share on Twitter Share on Google+
• bshifaw moonMember, Broadie, Moderator admin

  June 3

  I'll ask the dev team, in the meantime tell us more about the background. e.g. What's its origin (human, exome), what is the exact command you used, what was the sequencing platform, are you working with amplicon data?

  0 · Share on Twitter Share on Google+
• qiuqiu87 Member ❠

  June 4

  @bshifaw
  Thanks for your answer. The command I used is the default command with -O to output vcf. The data is about human sequence, the tumor detection panel with amplicon data. And as I paste the gvcf output, the qual of 32944609 is 0, so is filtered before output vcf.

  image.png 57.7K

  0 · Share on Twitter Share on Google+
• qiuqiu87 Member ❠

  June 4

  @bshifaw
  And I try some versions, and find 4.0.9.0 and higher version can't detect the insAAAA variants.

  0 · Share on Twitter Share on Google+
• bshifaw moonMember, Broadie, Moderator admin

  June 4

  I've asked the dev team, I'll let you know what they say.
  In the meantime let me know if using --adaptive-pruning and --kmer-size 35 as suggested in this forum post helps
  https://gatkforums.broadinstitute.org/gatk/discussion/comment/57464#Comment_57464

  0 · Share on Twitter Share on Google+
• qiuqiu87 Member ❠

  June 5

  Thanks， I tried both to use --adaptive-pruning true and to change the length of kmer-size 35, the variant in 32944609 is still missing in vcf. But I found in the gvcf output, the quality of 32944609 from 0 increase to 2451.05, but still filtered in vcf output.

  image.png 36.1K

  0 · Share on Twitter Share on Google+
• bshifaw moonMember, Broadie, Moderator admin

  June 6

  Then from this point the above message may apply

  The site in question may not have met the default threshold to be considered a variant by Haplotypecaller due to quality of evidence. However, the tool does have parameters that can be tweaked to be more sensitive to calling variants which can be found in the tool documentation page (warning, this increases the chance of false positives).

  0 · Share on Twitter Share on Google+