Germine copy number variation detection?

[etal]

1) What will be a good "reference"/normal BAM file? Randomly selecting reads from BAM files of some normal people sequencing data?

You can use a program like CoNIFER or cn.MOPS to process the samples as a cohort. These programs each use a statistical method to consider all of the given samples in aggregate to determine the "expected" baseline for copy number inference.

Alternatively, CNVkit works more like your proposal, where a reference copy number profile is initially constructed from a pool of normal BAM files and then reused to infer CNVs in individual samples. CNVkit can also operate without any normal-sample reference at all, but results will be slightly noisier.

2) Assume we have such a BAM files in 1), can we just use public software tools such as VarScan2 to compare the patient's BAM with the reference BAM?

Yes, you could. Be careful to check that any recurrent CNVs you find aren't simply the result of having a region of unusually high or low read depth in your reference BAM.

Most program for copy number inference will give you some indication of how the reference should be built and whether you should process one sample at a time or the whole cohort at once.

[mrfox]

Hi all,
I have the normal tissue of about 20 cancer patients with family cancer history. We would like to perform exome sequencing to identify germline mutations, especially germline copy number variations but have not much experience. A couple of questions:

1) What will be a good "reference"/normal BAM file? Randomly selecting reads from BAM files of some normal people sequencing data?
2) Assume we have such a BAM files in 1), can we just use public software tools such as VarScan2 to compare the patient's BAM with the reference BAM?

Any input is greatly appreciated!