Mendelian precison

Question

[heureuse]

Hello,

I would like to validate CNV using trios (father, mother and offspring) using mendelian precision. I'm not able to find a paper talking about the methodology in details. Do you know where I can find resources or how I can perform that type of analysis?

Thank you

[Pierre Lindenbaum]

"using mendelian precision"

what do you mean ?

[heureuse]

In the paper HadoopCNV(Yang H, Chen G, Lima L, Fang H, Jimenez L, Li M, Lyon GJ, He M, Wang K. HadoopCNV: A dynamic programming imputation algorithm to detect copy number variants from sequencing data.) they used mendelian precision to validate CNV.

They look at the CNVs inherited by the child vs the one that seems to be de novo. They used that approach to validate their tool (sensitivity and precision). They did not give details so I don't know how to approach it.

I was wondering if someone here know how to perform that types of analysis or have resources I can read.

Thank you

[Pierre Lindenbaum]

OK, so as far as I understand "mendelian precision" is "just" looking at the de-novo CNVs....

[heureuse]

Not for that but to know the % of CNV inherited in order to know the sensitivity of the caller

Answer 1

[Pierre Lindenbaum]

For trios and short-reads, I use delly or manta to get a join VCF of the trio (+ a few controls to discard false positives).

I then keep de-novo variants (using bcftools or another tool) .

Usually short SV or very large SV are false positive or are hard to validate.

A tool like duhold https://github.com/brentp/duphold can be used to filter-out false positives.

I the number of de novo CNV is low I use my tool swingbamcov https://jvarkit.readthedocs.io/en/latest/SwingBamCov/ to visualize and validate the CNVs

or you can try samplot : https://github.com/ryanlayer/samplot

[heureuse]

This is really helpful and beyond what I've expected. Thank you

[heureuse]

Hello, I look at these tools and they are really interesting. However I don't have bam files but gVCF files and VCF files. Do you know other approaches I can use to validate my SV ? I want to calculate the sensitivity and the precision. I've already validate with the genome from GIAB. I'm looking for a second way to validate (That's why I wanted to use mendelian precision).

Thanks