The effect of accidental selfing on the analysis of a diallel cross with sugar cane

Summary

The diallel cross was found to be an unsuitable design for quantitative genetic analyses with sugar cane.

Accidental selfing caused substantial inbreeding depression for most characters. Consequently, analysis of a diallel cross without regard to the effect of accidental selfing would be subject to bias, if accidental selfing were common. A selfing model proposed by Dudley (1963) was extended to include differential selfing, and was tested in an experiment in which the percentage of selfing in each cross was estimated visually.

Variance components had large standard errors, which prevented firm conclusions from being drawn, but selfing appeared to cause substantial bias in the estimates. Estimates of genetic variance components from a p(p-1) diallel set, using the extended selfing model, were compared with estimates obtained from the standard analyses in which reciprocal or maternal effects should account for the effect of selfing. The estimates were comparable, but the standard analyses were simpler and produced estimates with lower standard errors.

When the effects of selfing were minimized, general combining ability (g.c.a.) effects appeared to be much more important than specific combining ability (s.c.a.) effects for most characters.