Quantitative studies in two grain amaranth populations using two selection methods

Abstract

A population of Amaranthus cruentus (UC87) and another of hypochondriacus (UC99) were subjected to Sl (selfed) family analysis of genetic variation for plant height, plant weight, head length, head weight, threshing percentage, seed yield:height ratio, days to flowering, 500 seed weight, harvest index and seed yield per plant during the summer of 1982. Mass and recurrent Sl selections were initiated on the two populations for yield:height ratio and harvest index. These selections were advanced to second generation by mass selection in the summer of 1983. The Sl selection was also done for plant weight in UC87 and days to flowering in UC99. All generations of these selection experiments were evaluated together during the summer of 1984. Population UC87 had significant variation for threshing percentage, 500 seed weight and harvest index as indicated by Sl family analysis. High direct and indirect selection gains indicated presence of additive and additive x additive effects. This was confirmed for harvest index by component variance estimates. There were signs of overdominance in Sl selections. Population UC99 had significant genotypic variation for all traits except plant weight, head length and head weight. High direct and correlated selection gains indicated high additive and additive x additive effects. This was also confirmed by the genetic component estimates for yield:height ratio and harvest index. Realized gains from mass selection were higher than their expected values and the gains from Sl selection. Gains from the second cycle mass selection were lower than gains from the first cycle mass selection. These observations were attributed to the heterogeneity of UC99, which most likely involved competition, natural selection and higher homozygosity levels. In both populations mass selection was more efficient than S1 selection. This suggested a greater role of additive than the dominance component of variation. Based on correlation coefficients and multiple stepwise regression, the best yield predictors were plant height, head weight, threshing percentage and yield:height ratio. Harvest index was an important predictor for the yield of UC99 original population in 1984 only. Yield:height ratio was the most consistent predictor; the path-coefficient analyses indicated that most other variables influenced yield through yield:height ratio. Head weight was also an important indirect path for some predictors.