January 06, 2012

Epistasis and phantom heritability

A few years ago, I proposed the "lego-block" model of genetic variation. The main idea of this model is that the search for genes that "do" something is often misguided, because most genes don't actually do much of anything in themselves: they are commodity blocks (like lego pieces), and it is the way they are put together that influences the complex phenotypic outcome of an organism.

(By not "doing" something, I do not, of course, mean that they have no biological effect. Rather, I mean that they have no effect at a higher-order phenotypic trait, in the same way that the luminosity of pixels is irrelevant to the depicted image, it is rather the combination of pixels of different luminosity that produces an image)

I am not, of course, denying that there are genes with large positive/negative effects, but these traverse one of two possible trajectories during evolution:
  • Flicker: alleles of large negative effect arise, may persist for a few generations, but ultimately die out. They never amount to much of anything
  • Shine: alleles of large positive effect spread through the population quickly and become fixed
Alleles that flicker are much more common than alleles that shine. This is a simple consequence of the random nature of mutation: there are many more ways to break a system than to improve it, and the mutation mechanism is excellent in breaking down organisms: the survivors are the ones who carry a "manageable" mutation load, and, once in a very long while, adaptive alleles arise that quickly become fixed.

An alternative to the "lego-block" model of commodity alleles that produce positive/negative phenotypes due to the way they work together (epistasis) is the model of additive variation. According to this model, there is a plethora of genes of small positive/negative effect for a trait, and the final phenotypic expression is influenced by the sum of positive/negative alleles one inherits from their parents.

A new paper in PNAS provides strong evidence that the missing heritability is not due to our inability of finding loci of small effect, but rather to the fact that we've overestimated the heritability of traits. 

At the limit (a perfect "lego-block" world) there are absolutely no alleles that are individually associated with any traits. That does not mean that all individuals will be phenotypically indistinguishable! A great deal of phenotypic variation can still persist even in this case.

Here is a simple example:

A 10 0
B 0 10

The value of a trait depending on the alleles in two loci is shown, e.g., AC=10, AD=0, BC=0, BD=10.

It can be easily seen (due to symmetry) that whether one inherits A/B in one locus, or C/D in the other has no effect -in itself- on the trait. It is the combination of alleles that has a (huge) effect on the trait.

The paper is open access.

PNAS doi: 10.1073/pnas.1119675109

The mystery of missing heritability: Genetic interactions create phantom heritability

Or Zuk et al.

Human genetics has been haunted by the mystery of “missing heritability” of common traits. Although studies have discovered >1,200 variants associated with common diseases and traits, these variants typically appear to explain only a minority of the heritability. The proportion of heritability explained by a set of variants is the ratio of (i) the heritability due to these variants (numerator), estimated directly from their observed effects, to (ii) the total heritability (denominator), inferred indirectly from population data. The prevailing view has been that the explanation for missing heritability lies in the numerator—that is, in as-yet undiscovered variants. While many variants surely remain to be found, we show here that a substantial portion of missing heritability could arise from overestimation of the denominator, creating “phantom heritability.” Specifically, (i) estimates of total heritability implicitly assume the trait involves no genetic interactions (epistasis) among loci; (ii) this assumption is not justified, because models with interactions are also consistent with observable data; and (iii) under such models, the total heritability may be much smaller and thus the proportion of heritability explained much larger. For example, 80% of the currently missing heritability for Crohn's disease could be due to genetic interactions, if the disease involves interaction among three pathways. In short, missing heritability need not directly correspond to missing variants, because current estimates of total heritability may be significantly inflated by genetic interactions. Finally, we describe a method for estimating heritability from isolated populations that is not inflated by genetic interactions.


No comments: