What does the dominance term tell us when traits are highly polygenic?

ISG Workshop 2026 Day 3
1

Dominance is often described as something that occurs within a locus, eg. a non-linear difference in the trait for AA, Aa, and aa genotypes.

But when we fit an ADE model for a highly-polygenic trait with thousands of loci, what does “dominance” mean in that context?

Within a locus, dominance often has a biological reason (haploinsufficiency, gain-of-function, etc.). So when we’re talking about highly-polygenic traits with thousands of contributing loci, it seems weird to think that one may have a high proportion of SNPs that work that way and another would not.

Is this more of an “all traits have D, but for some traits we can’t look at it in a traditional twin model because C interferes” thing? (But I think I’ve seen AE models, so it probably doesn’t make sense to say all traits have D.)

With a highly polygenic trait, what is D actually telling us about the trait?

(And is D strictly for dominance between alleles in the same locus, or would it also pick up epistasis/GxG?)

Questions about the second day 3 video
2

OK- a few things. Terminology is confusing!

At a single locus, the parameter “d” the dominance parameter, quantifies the deviation from the midpoint of the two homozygotes. If there is biological dominance at a single locus, then “d” will be != 0.

This concept is different from the idea of “dominance variance” which can be defined several ways including as the genetic variance at a single locus, not explained by the additive action of alleles. Imagine a regression of (mean) phenotype on copies of the increaser allele at a single SNP. In other words, we have three phenotypic means- one for each genotype (AA, AB, BB), and we regress on number of copies of B. The variance explained by this regression is the additive genetic variance and the residual variance is the dominance genetic variance.

At a single biallelic locus, the additive genetic variance is

V_A = 2pq[a+d(q-p)]^2

And the dominance genetic variance is

V_D = (2pqd)^2

Confusingly, you’ll see the “d” dominance parameter contributes to V_A! In other words, dominant gene action shows up in the additive genetic variance component! V_D on the other hand is only > 0 when d != 0.

The additive genetic variance is the summed across all the contributing loci in the genome. Likewise for the dominance genetic variance. These are the components of variance that we (in principle) estimate in the classical twin design when we fit an ADE model.

That’s a long way of saying that we can say very little about gene action at a single (or multiple loci) on the basis of the classical twin model! D > 0 implies that there must be interactions between alleles (or loci) for at least some of the loci contributing to the trait.

In the case of epistatic variance (which arises from interactions *between* loci), this will manifest as “D” in the classical twin model

Check out these resources for a more in depth explanation of all of the above:

https://www.youtube.com/watch?v=1lR09FDVE5I

Hope that is helpful

Dave

1 Like