Choice of method for identifying germplasm with superior alleles. 2. Computer simulation results

An accurate and efficient method of screening the many germplasm sources available for their ability to improve elite, adapted germplasm is needed. The superiority measure (SX) of a population (P) was defined as the product of the frequency and relative superiority of the alleles in P that are more favorable than the best in an elite, adapted reference single cross I1×I2. A computer simulation was done to determine the correlations between various screening methods and the SX. The genetic model used included multiple alleles, no linkage, two types of non-epistatic gene action (additive and complete dominance) and two types of epistatic gene action (complementary and duplicate). Genetic variances in the populations and a statistic proposed by Dudley (SD={[P x I1-I1] [I1 xI2-I2]- [P x I2-I2]-[P x I2-I2] [I1 x I2-I1]}/{2[I1-I2]{) were inconsistently correlated with the SX over all types of gene action on the basis of rank correlations. The testcross to the single cross (TC[SC]=P x [I1 x I2]) and the upper bound on the SX (UBND=minimum [P x I1-I1, P x I2-I2]) were both consistently highly genetically correlated with the SX. In the set of populations simulated, there were positive correlations between products of allelic frequencies and effects at different classes of loci. The UBND usually had a higher rank correlation coefficient with the SX than did the TC(SC). The differences between their correlation coefficients were often insignificant. Although the TC(SC) gives no indication as to which inbred the population is more closely related, its ease of use and expected lowers standard error compared with the UBND indicate that it would be an appropriate choice of screening method for identifying superior populations in the sense defined.