Merging CNVs from multiple tumor samples from the same patient

Question

[sey]

I have multiple metastatic samples from the same cancer patient. I have used FACETS to call CNVs for each sample. Now, I would like to merge the individual results in a table where each row is a CNV, and each column is a sample.

Because different samples have different segments, I merge the samples by introducing breaks to the segments for each position where there is a CNV start or end in another sample. Example:

sample 1 start-end : copy number

1-100 : 3

101-200 : 1

sample 2 start-end : copy number

1-50 : 4

51-200 : 0

merged start-end : copy number(S1) : copy number(S2)

1-50 : 3 : 4

51-100 : 3 : 0

101-200 : 1 : 0

My question is: Instead of creating artificial breaks, it would be great to be able to narrow overlapping segments from different samples to common CNVs. Are there any previous examples (or tools) where this is done? Or, are there allele specific copy number tools which work with multiple samples, and do clonal heterogeneity analysis?

Answer 1

[markus.riester]

I think superFreq (https://github.com/ChristofferFlensburg/superFreq) is pretty much the only allele-specific caller designed for multiple samples. In case your data is from whole-genomes, there might be others.

The benefit of supporting multiple samples is mostly in tracking cellular fractions of SNVs, not so much in CNVs. CNVs are obtained from multiple data points and you thus get fairly accurate estimates; cellular fractions of SNVs have huge confidence intervals and then multiple time points help quite a bit. It can help significantly for CNVs when some time points are very noisy or of low purity.

Treeomics (https://github.com/johannesreiter/treeomics) might me relevant too.

Your approach with FACETs seems fine to me. The merging of segmentations is already implemented in the CNTools Bioconductor package, btw.