MODELING THE GENETIC BASIS OF HETEROSIS: TESTS OF ALTERNATIVE HYPOTHESES

Houle (1994) showed that marker-associated heterosis due to general inbreeding depression could not be distinguished from direct overdominance at the marker locus by examining mean genotypic fitnesses, in the one-locus case. Indeed, both hypotheses equally fit the same regression model, referred to as the "adaptive distance model" (Smouse 1986). I here extend the analysis to several loci and to the relationship between marker genotype and variance in fitness. Several predictions differ between the overdominance and inbreeding hypotheses: (1) all locus-specific effects are equal under inbreeding, whereas they are not under overdominance; (2) the adaptive distance model has an increasingly low fit when the number of loci increases, under inbreeding, whereas it always explains the whole variance in fitness under overdominance; (3) a negative relationship is predicted between mean fitness and the variance in fitness, under inbreeding, which is not predicted under overdominance. Some statistical tests are derived from these predictions, that help to identify the genetic basis of heterosis. Simulations show that the power of these tests allows their application to real datasets.