A general method for the detection and estimation of additive, dominance and epistatic variation for metrical traits

Summary

For plant species in which a single cross pollination yields one or few seed and self pollination is easier, the families of the rather demanding triple test cross design can be replaced by their progeny families produced by selfing. The theoretical consequences of making this substitution on the sensitivity of the tests for epistasis and on the reliability of the estimates of the additive and dominance components of variation in the presence and absence of epistasis are described for F₂ and F_∞ populations. The practical consequences are illustrated by comparative analyses of normal families of a triple test cross and their selfs for the F_∞ population using a population of 80 F₁₃ families produced from a cross between varieties 1 and 5 of Nicotiana rustica.

Both in theory and in practice the additive genetic component D is detected and estimated with similar reliabilities from the normal families of a triple test cross and their selfs. On the other hand, the dominance component H is detected and estimated with much lower reliability from the selfs and in the presence of epistasis the estimate is more biased. The sensitivity of the standard test for epistasis is also lower for selfs but for an F∞ population this is partially offset by the additional test for epistasis involving the means of the F_∞ families (P̄i) themselves.

In the comparative analyses family size and replications for the normal families and their selfs have been kept the same to make direct comparisons possible. In normal circumstances, however, one would not devote the time and effort required to produce the selfs unless as a result larger families and more replicates could be raised. Depending on the magnitude of the resulting increase in the number of plants raised the loss of reliability in the estimate of H and of sensitivity in the test for epistasis would be partly or wholly restored.