RFLP GENETIC SIMILARITY ESTIMATES AND COEFFICIENT OF PARENTAGE AS GENETIC VARIANCE PREDICTORS FOR SOYBEAN YIELD

RFLP genetic-similarity estimates (RFLP-GS) and coefficient of parenta ge (CP) have been used as measures of genetic similarity within crop s pecies. However, practical application of these measures in plant bree ding remains uncertain. This study was designed to probe the utility o f RFLP-GS and CP in predicting genetic variance (GV) for seed yield am ong inbred soybean [Glycine max (L.) Merr.] lines. To achieve this goa l, five single seed descent populations were studied, representing a r ange of RFLP-GS and CP between the parents from 57 to 75% and 0.06 to 0.5, respectively. The GV for yield was estimated for each population through field evaluation of 30 inbred lines per population, in two Nor th Carolina field locations during 1994. Both RFLP-GS and CP correctly identified the population with the highest GV; however, CP predicted GV for yield more efficiently (r(CP.RFLP-GS) = 0.91; r(CP.GV) = -0.81 ; and r(RFLP-GS.GV) = -0.58). The GV was near zero when the CP betwee n parents was larger than 0.27 or when RFLP-GS was larger than 75%. Ne ither genotype x environment interaction nor low field precision were factors for the lower predictive value of RFLP-GS. Expected gains from selection agreed partially with RFLP-GS results but closely matched C P and the actual fate of populations in a USDA breeding program. These results indicated that caution should be taken in an applied soybean breeding program when crossing parents with a relationship larger than half-sib or when the RFLP-GS is larger than 75% when yield improvemen t is the main breeding objective.