Clinical utilization of NGS data in cancer research

Question

[Max Ivon]

Greetings everyone!

I wonder one question which i didnt manage to google. NGS is actively used for cancer research, but is there any rule or recommendation for the allele frequency cutoff of the somatic mutation detected by NGS data to be clinically significant like in case of sanger sequencing (where mutation is supposed to be clinically significant with allele frequency of 15 or higher as i know). As i know there should be sufficiant amount of cells hurboring driver mutation for it to be targetable by drugs or any other invasion. Or in case of NGS statistical significance is enough?

Thanks in advance for the answer

[Sean Davis]

Are you using the data for making clinical decisions, or simply to find potential clinically interesting results for research?

[Max Ivon]

For clinical decisions - find potentially targetable mutation list of which is known.

[Sean Davis]

Then, you'll probably want to report everything that you can to the clinician so that he/she can make appropriate decisions. To my knowledge, there are no evidence-based "numeric rules" about what is clinically relevant.

Answer 1

[Sean Davis]

Statistical significance is a first step (in the sense that a somatic variant caller us used). Beyond that, one needs to determine whether the finding is a false positive (which goes beyond simple statistical significance to examine variant context and other variant databases). Biological significance is yet another level of interpretation and usually involves a combination of annotation and literature searching. Finally, one can begin to interpret clinical relevance. All of these steps present general challenges and each has challenges that are specific to disease and patient. That is why many groups use a molecular tumor board and/or formal clinical signout rather than relying strictly on numeric results.

In short, there is no simple answer to your question.

[Max Ivon]

Tank's for the answer, but don't you mind to tell me what can I start from to get deeper in this theme? I've tried to inspect TCGA mutation calling process, but haven't found anything useful except links to mutect

Answer 2

[poisonAlien]

Its hard to say. Many clinical significant variants do occur at low vaf's (subclonal). You can check this paper, where known drivers in AML such as DNMT3a or RUNX1 occur at <10% vaf. Again, if your depth is high enough, you could use statistical significance as first step to proceed.