SOURCES AND BIASES OF THE PREDICTORS OF THE PROPERTIES OF RECOMBINANT INBREDS PRODUCED BY SINGLE SEED DESCENT

Four sources of estimates of the additive genetic component of variance, whose square root (D) is one of the three parameters used to predict the properties of random inbreds derived from an F2 by single seed descent are compared theoretically and experimentally using estimates from two Micotiana rustica experiments. The comparisons are mainly based on the level of agreement between /D estimated from F2 and first backcross (B1 and B2) families and the L2i (F2 x P1), L2i (F2 x P2) and L3r (F2 x F1) families of an F2 triple test cross and that observed among the random inbreds and to a lesser extent on the magnitudes of the standard errors of the estimates. On level of agreement, the estimates from all sources were equally good in the absence of non-allelic interaction. In its presence those from the F2, B1 and B2 families were always among the worst. On balance the estimates from (L1i + L2i) families were best in the presence of duplicate interactions, those from (L1i + L2i + L3i) families were best in the presence of complimentary interactions and those from La families were best with a mixture of the two kinds of interaction. For estimates based on the same total number of individuals those from (L1i + L2i) families had the smallest and those from F2, B1 and B2 and Lsi families the largest sampling errors in most circumstances. © 1979 The Genetical Society of Great Britain.