Prediction criteria of promising F3 populations in durum wheat: A comparative study

Effective selection of parental material and promising segregating populations is an essential requirement for breeding success. There are many contradictive reports about the best parent selection criterion for the development of promising crosses. For the clarification of this problem field experiments were conducted for four consecutive years to compare the effectiveness of six criteria for the prediction of the most promising F3 populations in durum wheat (Triticum durum L.): the mid-parent value, the F1, the F2, the (F1 + F2)/2, and the genetic distance among the parents as it is calculated using the SSR and RAPD molecular markers. During the first growing season (2003-2004) nine commercial cultivars of durum wheat and four landraces were crossed. The following growing season (2004-2005), 17 crosses (F1 generation) were evaluated under low plant density (1.15 plants m−2) in a replicated (R-21) honeycomb design. During the third growing season (2005-2006), the four highest yielding crosses, one cross with an intermediate yield, and the three crosses with the lowest yield (F2 generation) were evaluated under low plant density in a R-9 honeycomb design. Finally, in the fourth growing season (2006-2007) progeny of the aforementioned eight crosses (F3 generation) and the ten parents were evaluated in a randomized complete block design in two locations. Furthermore, the genetic distance among the parents was determined using the SSR and RAPD molecular markers. It was observed that the three F3 populations with the lowest yielding ability were the ones with the lowest mid-parental value. In addition, one of the two top F3 populations was second in the rank according to the mid-parental value. Furthermore, the two top F3 populations were also the highest yielding in the F1 and F2 generations. On the contrary, none of these crosses were predicted by the genetic distance as it was calculated using the SSR and RAPD molecular markers. It was concluded that parental pairs with high mid-parental value and high combined yield (F1 + F2)/2 obtained after evaluation of their F1 and F2 at low plant density was the most effective way to predict promising F3 populations.