GATK VariantEval no concordance between dbSNP and our data

Answered

Follow

Pierre_Bioinfo

• May 03, 2021 07:50

Hi, 

I use VariantEval to calculate the percentage of variant present in dbSNP. So when I use VariantEval, with CompOverlap I have this result in nEvalVariants:

Novelty    nEvalVariants

  all              28611

known         27943

novel           668

But in my VCF sample, I have 106137 variants. So my question is: Why do I have 28611 variants in VariantEval and 106137 in my VCF file ? Normally "all" line must be correspond to the total number of variants of my vcf, right ?

Any thoughts appreciated !

Thank you!

0

• 

  Genevieve Brandt (she/her)

  • May 20, 2021 20:36

  Hi Pierre_Bioinfo,

  There are most likely many variants in your VCF that do not pass variant filtering, so they are not included in the "all" category yet they are in the file. You can check this by running SelectVariants with --exclude-filtered true and see how many variants are in the output file. Filtered variants will have something other than PASS in the FILTER column.

  Hope this helps!

  Best,

  Genevieve

  0

  Comment actions Permalink
• 

  Pierre_Bioinfo

  • July 12, 2021 13:08

  Thanks for your answer Genevieve !

  I just test your solution and I have same value with SelectVariant and my total number of variants in my VCFs ... I don't really understand why I have this results because I don't see any argument that can could do this and I use my last VCF ( therefore containing the best quality variants) to execute EvalVariant. You don't downsampling, right ?

  Best,

  Pierre

  0

  Comment actions Permalink
• 

  Genevieve Brandt (she/her)

  • July 12, 2021 18:16

  Hi Pierre_Bioinfo,

  Thanks for following up! I'll see if I can get to the bottom of this. First, could you share your VariantEval command? Also, what mode are you referring to with CompOverlap?

  Thank you!

  Genevieve

  0

  Comment actions Permalink
• 

  Pierre_Bioinfo

  • July 13, 2021 07:02

  Hello, 

  That's my command: 

  /path/to/GATK(4.2.0.0) VariantEval --eval /my/VCF/file -D dbsnp_146.hg38.vcf.gz -L bedFile.interval_list -no-ev true -EV CompOverlap -EV TiTvVariantEvaluator -O ${sample}_variantEval.txt

   

  I only use dbsnp stratrification and I use VariantEval to calculate the percentage of variant referenced in dbSNP and the Ti/Tv ratio.I don't quite see what you mean by "mode", so I hope I answer your question.

   

  Thanks,

   

  Pierre
  0

  Comment actions Permalink
• 

  Genevieve Brandt (she/her)

  • July 13, 2021 17:36

  Hi Pierre,

  Yes that answers my question, thank you! Could you give me more information about your interval list? VariantEval will only run for the variants that are within the interval list, so that could be causing fewer variants to show up in your VariantEval results.

  You can check this with SelectVariants using the same interval list. Let me know what you find.

  Best,

  Genevieve

  0

  Comment actions Permalink
• 

  Genevieve Brandt (she/her)

  • July 13, 2021 17:37

  Hi Pierre,

  Yes that answers my question, thank you! Could you give me more information about your interval list? VariantEval will only run for the variants that are within the interval list, so that could be causing fewer variants to show up in your VariantEval results.

  You can check this with SelectVariants using the same interval list. Let me know what you find.

  Best,

  Genevieve

  0

  Comment actions Permalink

Please sign in to leave a comment.