Evaluate the sensitivity and specificity of vcf files

Question

[jackycsie]

Hi,

I have already finished the variant call of NA12878.

My steps are as follows:

1. bwa mem
2. SortSamSpark
3. MarkDuplicatesSpark
4. BaseRecalibratorSpark
5. ApplyBQSRSpark
6. HaplotypeCaller

Now I have a vcf file, but I don't know how to judge his correctness.

My reference data is:

1. dbsnp_138.b37.vcf.
2. Mills_and_1000G_gold_standard.indels.b37.vcf
3. 1000G_phase1.indels.b37.vcf

Thank you, jacky.

[jackycsie]

Hi,

I find this url:

https://ftp-trace.ncbi.nlm.nih.gov/giab/ftp/release/NA12878_HG001/NISTv3.3.2/GRCh37/

It's right or only provide chromosome 1 ?

Thanks, Jacky.

[Mark]

I think this is better suited for the GATK forums. There's this post that might be a good starting point: https://gatkforums.broadinstitute.org/gatk/discussion/6308/evaluating-the-quality-of-a-variant-callset

[Jorge Amigo]

although chr1 is one of the biggest ones, and considering that you've already done the largest effort which is calling variants on NA12878, I'd definitely go for the entire genome's numbers rather than being fine with a relatively small subset of statistics.

Answer 1

[Jorge Amigo]

you've chosen a reference sample to apply your variant calling pipeline to. well done, because you're almost there. just get a set of high confidence NA12878 variants such as the GiB project ones, use a comparison tool such as RTG vcfeval (the one recommended by the GiB project), and you'll be done.