Optimization of PLS model of yield traits and quality traits of hulless barley in qinghai-tibet plateau

In order to construct and continually optimize the mathematic model of traits such as plant height, grain number per spike (GNS), grain weight per 1000-grain weight and trait factors such as lysine and β-glucan and to reveal internal relations between yield traits and quality traits. Using partial least squares regression (PLSR) method, the yield and quality traits of 83 highland barley collected from different ecological planting areas of Qinghai-Tibet Plateau were studied by research on optimization of mathematical model. The results showed that the yield and quality traits of PLS (consider the interaction) model has the highest accuracy. The factors affecting the yield traits of lysine followed the order of spike length > spikelet density > 1000-grain weight > plant height > grains number per spike;yield traits affecting starch were acoorded with the sequence of spike length > spikelet density > grains number per spiker > spike length and spike density interaction > plant height > spike length and 1000-grain weight interaction. The factors affecting the yield traits of β-glucan were as follows: spikelet density, spike length, plant height, 1000-grain weight, spike length and spikelet density interaction; Yield traits affecting protein abided by spikelet density > 1000-grain weight > grains number per spiker > spike length > plant height. The factors of yield traits such as spike length, spikelet density and 1000-grain weight had significant effects on the quality traits of lysine and β-glucan in highland barley, but combined effects. In summary, quality traits of the highland barley are the result of the comprehensive effect of each yield trait factor. This study provides a theoretical basis for guiding good quality and high yield breeding and cultivation of highland barley. (authors)