CNV detection

Question

[heureuse]

Hello all,

I saw a post about detecting CNV from VCF file. People advised to avoid doing that and use a bam file instead.

My concern is I think that VCF or gVCF can be a good alternative when we have large cohort (400 000 sample for example). Algorithms based on BAM files consume a lot in terms of time and computational resources.

What do you think about that ? Thanks

Answer 1

[cmdcolin]

VCF files, just plain SNP VCF files for example, can provide several "signals" of information for detecting CNVs including

• calculating the B-allele frequency ("BAF") from e.g. a normal SNP VCF, and plot this value or use it as input to a CNV detection

see https://cnvkit.readthedocs.io/en/stable/baf.html https://help.dragen.illumina.com/product-guides/dragen-v4.3/dragen-dna-pipeline/small-variant-calling/baf-output

• calculating the read depth with the DP field from a normal SNP VCF

you can get several results looking up "detect CNV from VCF", and indeed, many pipelines like (from cursory looking: cnvpytor, Illumina/canvas) will use both BAM and VCF for adding the B-allele frequency information

Answer 2

[andrew]

Vcf and gvcf are more efficient than bam files