how to select gene mutation

Question

[1561775911]

Hi,biostars,

I have 200 patients with whole exome sequence and I want to build a multivariate Cox regression of overall survival with the gene mutations. First of all ,I should choose the mutational genes for the next modeling , while I don't know how to set threshold of mutation frequency to choose them. It is said that 1% is the putative threshold above which mutation is considered as 'high frequency mutation'. But if I choose 1%, it means certain gene is mutated within 2 patients and 198 patients without this mutation. In this case, is the gene statistically appropriate for Cox regression analysis?

Thus,I hope someone could give me some advice on the mutation frequency threshold setting or articles which deal with similar situation. Thank you!

---

[Emily]

I have edited your tags, splitting them into two. People on BioStars use tags to find posts, so having easily searchable tags can help to get your questions answered. When inputting your tags, hit return/enter after each tag to make them come up separately. Each topic should be its own tag. Suggested separate tags are: the topic you're looking at (mutation frequency), the tool or analysis you're trying to use (cox regression) and any relevant programming language.

[Kevin Blighe]

It is said that 1% is the putative threshold above which mutation is considered as 'high frequency mutation'.

Where is that said?; which web-site or manuscript have you been reading?; are you sure that 1% does not relate to the mutation being mutated in >1% of the tumor clones in a tumour bulk biopsy?

[1561775911]

I have asked a Phd of genetics and he told me 1% was a custom ,while there is no standard for the mutation frequency threshold.That is why I ask if there is anyone who has more experience about this

Answer 1

[Kevin Blighe]

Well, there you go: ask 2 people (even Professors in Statistical Genetics), and they will tell you 2 different things! The threshold can be modified according to the dataset at hand. A naïve person would be one who rigidly adheres to one threshold over another. Even 1% could be considered extremely high, as, in a population of 1 million people, 1% is equivalent to 10 thousand people, which is an entire Irish town.

Also, remember that even variants with 'high' allele frequencies can be functional and increase risk of disease. For example, SNPs in the CCND1 locus have MAFs of >10%, yet, they are amongst the variants that are most highly statistically significantly associated with ER-positive breast cancer.

Conclusion: selection of functional / pathogenic variants should be neither solely nor heavily based on a frequency filter.

Further reading: Rare and Common Variants: Twenty arguments

Kevin