Computer simulation of family selection schemes suitable for kale (Brassica oleracea L.), involving half-sib, full-sib and selfed families.

Three recurrent selection schemes suitable for kale (Brassica oleracea L.), involving half-sib (HS), full-sib (FS) and selfed (S) families, were compared by computer simulation. All combinations of 6, 12 and 24 families selected, out of 120 and 240 assessed, were investigated for a range of genetical models. Selection was simulated for 20 generations from an initial allele frequency of 0.05 and for 16 generations from an initial frequency of 0.20. With an initial frequency of 0.05 there was a serious loss of desired alleles ranging from 0.31 out of 20 for the HS scheme with 24 out of 240 families selected to 9.19 for the S scheme with 6 out of 120 families selected. It was concluded that if as many as 20 cultivars were included in the initial population the selection scheme should be chosen to minimise the loss. With an initial frequency of 0.20 there were no losses with 12 and 24 families selected in the HS and FS schemes respectively, and the highest loss was 2.88 for the S scheme with 6 out of 120 families selected. It was concluded that if as few as five cultivars were included in the initial population a compromise between the initial response to selection and the loss of desired alleles should be sought. Selecting 6, 12 and 24 families for the HS, FS and S schemes respectively, resulted in average relative responses per generation of 2.28, 2.74 and 2.76, respectively for the first five generations, and losses of 0.22, 0.13 and 0.35, respectively after 16 generations. Practical considerations favour the FS scheme over the S scheme.