Simulation of quantitative characters by genes with biochemically definable action : IV. The analysis of heritable variation by the diallel technique.

Six groups of genetic materials were developed in a cruciferous garden plant, Matthiola incana R. Br., to produce simplified genetic systems of a pair of loci in each group. There were only two alleles at each locus, which were directly involved in the modification of anthocyanins in the plant tissue. The parental lines and their F1's in each group constituted an ideal 4 × 4 diallel cross and satisfied all but one condition necessary for a valid diallel analysis. Nonallelic interaction was the only possible offending postulate.The diallel analysis of the data on anthocyanin content in the flower tissues and a comparison of the results with that of a relatively straightforward method of analysis indicated that in the presence of epistasis, the dominance ratio (H1/D) ceases to be a reliable measure of the average degree of dominance. In such situations additional genetic information obtained from the diallel analysis are not in agreement with the expectations on the basis of the already available genetic information on the materials. The estimator H2/4 H1, a measure of average value of the product of alleles with positive and negative effects, seemed to remain unaffected by epistasis. The Wr/Vr regression analysis does not always permit the detection of nonallelic gene interactions. The results suggest that duplicate interaction may escape detection by the regression analysis.