How to identify CNV in Tumor-only samples?

Question

[jinxinhao1988]

Dear all,

This problem had been bothered me for a long time. Because some tumor samples can not find a normal sample to compare with. But you still want to analyze the CNV, what can we do about this?

My thoughts is to find some "stable genes" as control. Then comparing the target fragment with these control genes to get the CNV information. It's kind of the similar with beta-actin in the qPCR experiment. If this is a possible solution?

But the problem is:

1) How to evaluate some genes are "stable" across the samples?

2) Maybe we should make sure that the control genes are only one copy in each sample to make the comparison easier.

I hope that I can get some suggestions related to this issue. Thanks in advance.

Answer 1

[Kevin Blighe]

There are some previous answers for this:

• "Copy Number Variation" detection in Tumor only sample
• Tools For Analyzing Copy Number Variation On All-Tumor Exome-Seq Samples

The stable genes idea is nice but you face the same issue of not knowing for certain whether the genes are indeed exhibiting normal copy number. Remember that CNV, in fact, refers to normal / healthy variation in a genome. Each of our genomes varies based not only on SNPs but also on CNVs. CNA, copy number alterations, more relate to somatic copy number changes.

Beta actin (ACTB), for example, exhibits 'normal' CNV and is therefore not a reliable reference: http://dgv.tcag.ca/gb2/gbrowse/dgv2_hg19/?name=actb;search=Search

I think that, in the absence of a matched normal, most tools build some sort of reference model using all samples, and then fit some form of model, such as a Hidden Markov Model, in order to gauge gain or loss in individual samples.

Kevin