Diallel Analysis Models: A Comparison of Certain Genetic Statistics

Abstract

ABSTRACTDiallel analyses, although popular, raise key questions: (i) Do genetic statistics from different diallel methods with same and different sets of lines offer the same or similar information? and (ii) Would lack of information on reciprocal (REC) and maternal (MAT) effects in certain diallels put breeding programs at risk? To answer these questions, general (GCA) and specific combining ability (SCA), MAT, and REC effects were calculated from diallels with 12, 10, 8, 6, and 4 maize (Zea mays L.) parents. We found that with the same lines, the signs and magnitudes of GCA effects were similar and highly correlated among Griffing's Methods 1, 2, 3, and 4 and identical between Griffing's Method 4 and Gardner and Eberhart analysis III (G&EIII). The SCA effects from Griffing's Methods 1 to 4 were correlated. Griffing's Method 4 produced SCA effects identical to G&EII and G&EIII; SCA effects from Griffing's Methods 1 and 2 were upwardly biased because of inclusion of parental lines. Griffing's Method 3 (no parents included) was the best method among all six methods; it gave unbiased SCA estimates and information on MAT and REC effects. With different numbers of lines in diallel crosses, genetic statistics could serve as references across breeding programs because the signs of GCA and SCA effects were mostly same; however, caution must be exercised because some effects with opposite signs were detected. A diallel method that does not provide REC and MAT information could miss identifying superior hybrid(s). Because Methods 1 and 2 provided upwardly biased estimates of GCA and SCA effects, they are not recommended for estimation of GCA and SCA effects. We also concluded that Method 3 was the best diallel method; it provided unbiased estimates of SCA effects and also provided crucial REC and MAT effects, which are useful for proper selection of male and female parents in hybrid development.