A statistician wonders about the influence of additive variance.

Where complex problems are concerned, it makes things simpler if some factors can be safely ignored. In quantitative genetics, one such assumption is that the bulk of genetic variation is additive. That is, the effect of an allele — a particular version of a gene — can be adequately described by its average effect in a population. But we know that genes often do not act additively; alleles interact, both with others of the same gene (a phenomenon known as dominance) and those of different genes (epistasis). All this contributes to the total genetic variation. But does this matter?

This question is tackled by Hill et al. (PLoS Genet. 4, e1000008; 2008). Reviewing the literature, they show that additive genetic variance is often close to total genetic variance. The authors then look at some mathematical models with strong non-additive genetic effects, and average over reasonable distributions of allele frequencies to show that the genetic variance is mainly additive. So non-additive genetic variation is usually of minor significance and we can continue to concentrate on additive genetic variance.

This is probably true on average, but may not always be so. Any trait is affected by only a finite, and in some cases small, number of genes. So averaging over all possible allele frequencies may say little about a particular case. There is also a much subtler problem. The authors conclude that additive genetic variance swamps other types of variation largely because most alleles common to a population occur with close to 100% frequency. But these extreme frequencies also reduce the total genetic variance. So, in practice, a lot of traits with strong additive effects might be classified as having no detectable genetic variation, and overall the importance of additive genetic effects would be diminished. Is this a genuine problem? Ah, more research is obviously needed.

Discuss this paper at http://blogs.nature.com/nature/journalclub