The Utilization Of Near Infrared Reflectance Spectroscopy In Improvement Of Rice Cooking And Nutrient Traits

Rice is one of the main cereals in China. Currently the emphasis in rice breeding basing on yield is being shifted towards quality improvement. The cooking and nutrient quality traits of rice contain amylose content, gel consistency, protein content and so on. Although there were some studies conducted on rice quality improvement based on the use of NIRS technology, there were still many limitations. For example the source of samples used were relatively homogeneous; most of the materials were from one year trials; sample quantity for the calibration set was less than required and the practical relevance of the models constructed were also not ideal. Even the current calibration model trquires a larger sample size of intermediate breeding lines. And most importantly, almost all samples in model were stable varieties or lines, which have a different separation status from that of intermediate lines and the sample status is quite important in NIRS analysis.In present research, the relatively large differences in cooking and nutrient traits of rice were studied by analyzing the seed composition of the original population within the same growth period of 9 rice cultivars including Z03-423, Fu 137, Zhefu 0515, G 04-44, Zhe 508, Zhenan 3B, Zhenongda 104, Saishu and Zhenong 952, which were selected as parents for achieving the separate breeding lines between 2005 to 2008. Using Foss NIRSystems 5000-C, the near infrared reflectance spectroscopy (NIRS) model was established to determine the cooking and nutrient traits including amylose content, gel consistency and protein content of rice, which could be used to appraise the rice germplasms or breed quality seeds. It constitutes a simpler and effective way to ensure an accurate screening within the early generations to help speed up the breeding process and can also serve as a useful analysis method for related studies. The main results were as follows:1. By using the CENTER software for defining the validation sets, 1176 samples for cooking quality traits including amylose content, gel consistency and 364 samples for protein content of rice were screened. The results showed that without determining the chemical constituents, just the principal component analysis (PCA) method could be used to select samples. However, selecting the appropriate values for the sample composition using Neighborhood (NH) analysis will help reduce the workload of chemical analysis. When using different samples, options could be compared and selected based on the actual prevailing conditions.2. The results from different spectral pretreatments of principal components for the creation of NIRS calibration models were significant. In rice, the results from the treatments of the first derivative by the combined use of spectral and scatter correction in the normal validation of NIRS pre treatments for amylose content and gel consistency traits using NIRS models were good; and so were the results from the second derivative.3. Within the whole spectral range (1100～2500nm), there was a negative correlation between amylose content and gel consistency, suggesting that these two traits were affected by same chemical constituents. The results showed the emergence of 6 wavelengths with high correlations for gel consistency, which were 1510, 1700, 2100, 2226, 2276 and 2378nm. There were radix overlapping frequency absorptions at 2096-2106nm.4. Among the rice grain, brown rice, milled rice and powder of milled rice, the analytical NIRS results obtained from brown rice and milled rice were better than that of rice grain which was suitable for analyze amylose content or protein content analysis. NIRS models for amylose content and protein content of rice using 3g and 0.5g of brown rice were better than that for gel consistency. The regression square of NIRS increased and standard error decreased when 3g of milled rice or smaller amount of fine rice powder (0.5g) were used. These NIR models could be applied in rice quality breeding for screening the rice plants with better amylose content, gel consistency or protein content.5. There were obvious differences in the validation of different populations and the differences in years affected rice cooking and nutrient quality traits in NIR model analysis. At the same time, the different components in the samples were also variable. NIR analytical error between the years could be eliminated or reduced to minimum by increasing the variability in the samples.