Allele frequencies

[bettywong]

Hello everybody,

 

I'm new to this forum. I was searching online where I could ask a genetic question, as I got stuck in my work...then I found here....any opinions are much appreciated...

 

Okay, jump into the problem...I'm doing a genetic association study, and considering to combine 2 East Asian population (i.e., Japanese and Taiwanese) as one to somehow increase the sample size, thus improve the power of the study. But I found out the gene we are looking at showed different allele frequency distributions in Japanese and Taiwanese, e.g., A-allele frequency is 0.48 and a-allele freq. is 0.52 in Japanese; but A-allele freq. 0.56, a-allele freq 0.44 in Taiwanese for the same gene.

 

Can I still combine them as one Asian group for analysis, or I should consider population substructure seriously in this case, thus analyze them seperately?

 

Thank you for your help..

[Mr Skeptic]

Can I still combine them as one Asian group for analysis, or I should consider population substructure seriously in this case, thus analyze them seperately?

 

Yes.

 

Which you want to do depends on what you want your study to be about. Do you want a larger study about Asians with very poor sampling except for two of the Asian subgroups? Or two smaller studies about two different Asian subgroups? I imagine you're better off doing each study separately and then combining them together if the answers match. The data collection would only require a little more effort, and since the calculations will all be the same that too will require little enough extra effort, and you get more results doing them separately. Just be sure to check whether the certainty level is high enough.

[bettywong]

I imagine you're better off doing each study separately and then combining them together if the answers match.

 

Thank you, I agree

 

Just be sure to check whether the certainty level is high enough.

 

You mean if the two results highly matched?

[CharonY]

It depends very much what you want to test, of course. If, for instance you want to test difference in allele frequencies in two populations, merging two subgroups are going to increase the variance, and hence reduce the sensitivity of the test. On the other hand, it may be important, not only for the sake of statistical power, but also for the kind of hypothesis that you are forming, to combine them.

[bettywong]

It depends very much what you want to test, of course. If, for instance you want to test difference in allele frequencies in two populations, merging two subgroups are going to increase the variance, and hence reduce the sensitivity of the test. On the other hand, it may be important, not only for the sake of statistical power, but also for the kind of hypothesis that you are forming, to combine them.

 

 

The hypothesis is to test any difference between cases and controls (i.e., genotype frequency difference) in Asian population, including subjects from Japan and Taiwan (both having their own cases and controls). So, I think I better analyze them seperately given the problem I have?

 

May I ask, under what kind of hypothesis, that you could combine them? Couldn't think of an example by myself, thanks indeed.

[CharonY]

There are a number of possible scenarios, one is extensive exchange, another one would be the prevalence of a given strong selective pressure, which is absent in the other population(s) and so on. Any population boundary is to a given extent arbitrary and should be used to best reflect the hypotheses to be tested. For instance, nations appear to be a good delimiter. However, depending on the geographic size and distribution certain sub-populations are very likely to cross borders. Depending on the way samples were obtained, bias may be introduced. Islands tend to be easier, but then Japan has quite a few of them and so on.

From the way the question is outlined it appears that there is also the risk of multiple hypothesis testing here (just as a sidenote).

[bettywong]

There are a number of possible scenarios, one is extensive exchange, another one would be the prevalence of a given strong selective pressure, which is absent in the other population(s) and so on. Any population boundary is to a given extent arbitrary and should be used to best reflect the hypotheses to be tested. For instance, nations appear to be a good delimiter. However, depending on the geographic size and distribution certain sub-populations are very likely to cross borders. Depending on the way samples were obtained, bias may be introduced. Islands tend to be easier, but then Japan has quite a few of them and so on.

From the way the question is outlined it appears that there is also the risk of multiple hypothesis testing here (just as a sidenote).

 

yes..there are secondary hypothesis to be tested in the cohort..the significance level will be corrected according to the number of tests

 

So...to summarize, if I understood properly from your posts, this difference (allele freq.) is somehow expected given sampling issue across whole Japan (consisting of several islands). To combine or not, depends on what I want to answer in my study given different samples I have.

 

For example, 2 genes were investigated in Japanese and Taiwanese, the main hypothesis is to see any difference existing in cases and controls in Asian population. One gene showed quite similar allele/genotype freq. in the controls of the 2 populations, I think I could combine them; for the other gene, which showed differed distribution in the 2 populations, may be due to positive selection or sampling bias, I should do seperate analysis directly.

 

Would you agree?