Somatic mutation calling without matched normal

Question

[Noushin N]

Dear everyone,

I have been asked to comment on an experiment design that involves WES or WGS of cancer cell lines which lack matched normal.

I know that this design is far from ideal, but I was wondering if there are people who have already stream-lined it (to the extent possible of course). I can think of a couple of common variation filters to keep out the germline variants, but I am not sure how to go beyond.

I am sorry if this is a duplicate question; I couldn't find a related hit.

Thank you,

Noushin

[Christof Winter]

See also Discrimination Between Germline And Somatic Mutations In Tumor Without The Availability Of The Normal Paired Sample

Answer 1

[Chris Miller]

The bottom line is that without a matched normal, you're just not going to be able to call the somatic status for the vast majority of sites. That said, you can winnow down a list to those you *suspect* are somatic. Some ideas:

• Weed out sites with high frequency in the population
• If your tumor is very impure, you can take advantage of the fact that the frequencies of somatic variants will be shifted away from 50%/100%

[Christof Winter]

Note that the variant allele frequencies will be shifted from 50%/100% even in a pure tumor such as a cell line if you have copy number alterations. Depending on the cancer type, cancer cell lines can easily be triploid or tetraploid, with many copy number aberrations.

[Noushin N]

That's a good point, but one is probably safe in assuming relatively high purity for cell lines (please correct me if I am wrong) compared to tissue biopsies. The issue about aberrant karyotype distorting allelic fractions is absolutely true, and needs to be accounted for.

[Christof Winter]

I think unless there are contaminating cells in the culture, your cell line DNA extracts should be close to 100% pure.

Answer 2

[Christof Winter]

Since you are asking about cell lines, maybe these have already been sequenced. There are bigger studies I'm aware of:

1. Cancer Cell Line Encyclopedia (CCLE), see http://www.ncbi.nlm.nih.gov/pubmed/22460905
   Browse and download the data: http://www.broadinstitute.org/ccle/home
2. NCI-60 cell line, see e.g. here: http://www.cbioportal.org/public-portal/study.do?cancer_study_id=cellline_nci60

Then I would check against COSMIC for known somatic mutations. And maybe the cell line in question even has data in COSMIC, see By Sample at http://cancer.sanger.ac.uk/cancergenome/projects/cosmic/

[Noushin N]

Thanks for pointing out these useful resources Christof. The cell lines in question are not previously characterized. Using COSMIC to identify known somatic mutations sounds good. But it seems to me it will be hard to call novel somatic mutations as they will be almost indistinguishable from private/rare germline variants, unless for those which are present at smaller cellular frequencies. Does this seem right?

[Christof Winter]

Yes, I would also say it's close to impossible to distinguish between novel somatic and private or rare germline variants that are not dbSNP/1000genomes/ESP6500. Looking at both mutation allele frequency and copy number at that position might help for cases where a somatic mutation occurred after a copy number change, but it's still rather guessing than identifying.