Question: Phenotype choice for one of a MZ twin pair in a genetic association study
gravatar for coleman_jonathan
5.5 years ago by
European Union
coleman_jonathan420 wrote:

(I don't know if this is really the right forum for this question, but haven't found a good answer elsewhere)


I was having a discussion with a colleague and this question came up...

It is common in molecular genetics studies to only genotype one of a pair of monozygotic twins (and then implicitly impute the other twin's genotype by assuming the pair is genetically identical).

My question is this: assuming a quantitative trait study, what should the phenotype be? Should it be the phenotype of the genotyped twin OR should it be an average of the two twins' phenotypes?


  • I genotype Twin A
  • Twin A has a phenotype score of 4
  • Twin B has a phenotype score of 6
  • Should Twin A's phenotype for the association study be 4 (the individual's phenotype) or 5 (the average phenotype of the genome, that is the average across the two genetically-identical twins)?

Answers from experience and from theory welcome!


phenotypes twins gwas • 1.4k views
ADD COMMENTlink modified 4.6 years ago by KJ10 • written 5.5 years ago by coleman_jonathan420
gravatar for KJ
4.6 years ago by
KJ10 wrote:

Even though this question is a bit old, I thought I'd answer this in case anyone someone else has a question like this.

Even though MZ twins are not 100% genetically identical, you can indeed assume the MZ twins to be genotypically identical when the data comes from an array. You should however never assume that the twins have the same phenotype! That could only be the case if the phenotype depends fully on genotype, which is never the case (they always have a slightly different environment).

If you can incorporate MZ twins in your model, use both twins with each their own phenotype. Otherwise, choose one at random, either the one you genotyped or the other since you can assume they are genotypically the same anyway.

ADD COMMENTlink written 4.6 years ago by KJ10

Thanks - so you would effectively take one phenotype at random, rather than averaging?

ADD REPLYlink modified 5 months ago by RamRS27k • written 4.4 years ago by coleman_jonathan420

Yes, but preferably using both (if it's possible to correct for relatedness in your study)

ADD REPLYlink written 4.0 years ago by KJ10
Please log in to add an answer.


Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 1373 users visited in the last hour