[SOLVED] P value of GenABEL when using permutation

Questions about GenABEL (aka *ABEL) suite of packages
Forum rules
Please remember not to post any sensitive data on this public forum.
The first few posts of newly registered users will be moderated in order to filter out any spammers.

When get a solution to the problem you posted, please change the topic name (e.g. from "how to ..." to "[SOLVED] how to ..."). This will make it easier for the community to follow the posts yet to be attended.
daisy8098
Posts: 2
Joined: Mon Feb 28, 2011 7:55 am

[SOLVED] P value of GenABEL when using permutation

Postby daisy8098 » Mon Feb 28, 2011 9:38 am

Hello!
I'm using GenABEL to perform GWAS. Since some traits of analysis are not in normal distribution, I want to use the permutation method. I used the formula "qtscore" and got the empirical P value. The question is for some traits which are in normal distribution there is no use permutation, and the results of P value are in form such as xxe-7, for the traits that used permutation, the empirical results are in form of such as 0.03. If I want to show the results in the same level, how can I get the true P value of each SNP and the value of threshold after permutation ?
I've read the R program of "qtscore", but not quite understood of it. I can't solve this problem and looking forward to your reply.

Thanks a lot,
Daisy
Last edited by daisy8098 on Tue May 03, 2011 8:01 am, edited 1 time in total.

yurii
GenABEL developer
GenABEL developer
Posts: 263
Joined: Fri Jan 21, 2011 5:20 pm

Re: P value of GenABEL when using permutation

Postby yurii » Thu Mar 03, 2011 9:36 pm

Hi Daisy,

I think there is some misunderstanding here -- 'permutation' p-values reported by GenABEL's qtscore, etc. using 'times' argument (or emp.qtscore, etc.) are experiment-wise empirical P-values, not "nominal empirical" P-values. In that, any 'permutation' P<0.05 is to be considered significant.

When you run it genome-wide, this is genome-wide P-values. If you run only a region, this would tell you region-wise significance.

So this is very much expected, that (say for normally distributed trait for simplicity) you do a GWAS with 300k, and if for a certain SNP you get nominal p value of say 1e-7, then

* after Bonferroni correction this corresponds to multiple testing adjusted P-value of 1e-7*300000=0.03

* after permutations procedure you get 0.02 or so

Hope this helps -- let us know!
Note that (Gen)ABELs are dynamically developing; while this post is intended to provide full information at the time of posting, please read on further posts, if any, as the topic may be updated with novel solutions at a later stage.

best regards,
Yurii

daisy8098
Posts: 2
Joined: Mon Feb 28, 2011 7:55 am

Re: P value of GenABEL when using permutation

Postby daisy8098 » Thu Mar 10, 2011 9:16 am

I think perhaps you think such expression is much intuitive when we used permutation. :)

The question confused me is I'm analysing about 10 traits. First I did the Shapiro test to check if they were normal distribution and found about half of them were not. For the normal distribution traits I used qtscore with times=1 and the rest used permutation method. I want to graphics display the results. There is no problem with the first part of the traits, but for the second part of the traits, the P value of most SNPs is close to 1. For example, I got 3 SNPs with P<0.05, 28 SNPs with 0.05<P<0.98 and the rest P=1. When I draw a picture by -log10 transform, 19 chromosomes were straight line and the distance between the significance SNPs and the P=1 SNPs is very large. I means that I want to show the results of the 10 traits together, can I get the original P value and the threshold for the permutation traits or how can I solve such problem?

I used BOX-COX transformation for such traits, but I think permutation is a better method.

Thank you.

yurii
GenABEL developer
GenABEL developer
Posts: 263
Joined: Fri Jan 21, 2011 5:20 pm

Re: P value of GenABEL when using permutation

Postby yurii » Thu Mar 17, 2011 12:55 am

Well, this is my personal opinion, and it can be argued, but if I was aiming to a manuscript I would have either used permutation procedure for all the traits, or no permutations for all the traits. For me it sounds like a more straightforward procedure to explain.

And if you use permutations to estimate genome-wide significance, you would not normally show the graphs -- they are ugly :)
Note that (Gen)ABELs are dynamically developing; while this post is intended to provide full information at the time of posting, please read on further posts, if any, as the topic may be updated with novel solutions at a later stage.

best regards,
Yurii


Return to “GenABEL”

Who is online

Users browsing this forum: No registered users and 2 guests