| Cotton is the one of most important economic crops in our country, which also takes a pivotal position in national economy and people’s livelihood. Cottonseed is the main by-product of cotton production, whose national output can reach10million tons or more, and the national output of cottonseed cake can reach600million tons. However, the nutrients of cottonseed cake have not be fully utilized due to the phytic acid which affect the nutritional value of cottonseed cake as animal feed. It is practically shown that the breeding and utilization of low phytic acid cotton cultivar is the most effective solution to solve this problem, and selection of low phytic acid cotton germplasm is the foundation of cotton breeding for low phytic acid in cottonseeds.At present, the phytic acid content can be determined by titration method, spectrophotometric method, high performance liquid chromatography (HPLC) method, and ion chromatography method etc. There are some disadvantages in sample preparation, analysis with high poisonousness, time-consuming and costly etc. for those methods. Fortunately, near infrared spectroscopy combining with multivariate calibration can solved those problems. The main objective of this work was to establish an accurate and stable calibration model for phytic acid determination in cottonseeds which may provide the foundation for cotton germplasm screening with low phytic acid content, based on NIR spectroscopy with suitable regression methods and variable selections. The main results were as follows:(1) Total304samples which were selected from the cotton breeding materials grown in different environmental regions and different years were used as the samples in this experiment. Using ion chromatography method determined the phytic acid content of cottonseeds. The results show that the distribution of phytic acid content is0.39～3.24%, and the average is of1.27%in304materials. This method overcomes the titration method’s poor accuracy, spectrophotometer method’s disadvantage of unstable, and has better separation effect than high performance liquid chromatography (HPLC) method. It was proved to be a high accuracy and feasible chemical method with high repeatability(RSD=2.28%), precision(RSD=0.57%<1%), and recovery rate can be applied to determining the phytic acid content in cottonseeds.(2) The spectra of cottonseed kernel powder and intact cottonseed were collected by NIR instrument, respectively. Then, the standard values of phytic acid content in304samples were determined by ion chromatography. Before developing the calibration models, standard normal variation (SNV) and Savitzky-Golay (SG) derivate were applied to spectra preprocessing. During developing the calibration models, we select several methods including Partial least squares(PLS), Least squares support vector machines(LS-SVM), Weighted least squares support vector machines(WLS-SVM) and Robust weighted least squares support vector machines(RWLS-SVM) in order to obtain optimal phytic acid calibration models of cottonseed kernel powder and intact cottonseed.(3) To the calibration models of cottonseed kernel powder, the predictive performance of nonlinear LS-SVM method was validated to be the best of all. The coefficient of determination (R2), residual predictive deviation (RPD), root-mean square error of cross-validation (RMSECV) and root-mean square error of prediction (RMSEP) were0.9176,3.4517,0.2143and0.1221, respectively. It showed that the calibration model of cottonseed kernel powder was achieved with good accuracy and robustness (R2>9.0000), and the models are able to substitute for usual method.(4) To the calibration models of intact cottonseed, the predictive performance of RWLS-SVM method was validated to be much better than the others. Its’ R2, RPD, RMSECV and RMSEP were0.7771,2.0258,0.2784and0.2471, respectively. The efficiency of this calibration model was not as good as that of the optimal calibration model of cottonseed kernel powder, and it is still needs to be improved.The calibration model of cottonseed kernel powder can be applied for phytic acid content determination in cottonseeds, providing a fast and efficient method of low phytic acid germplasm screening for low phytic acid cotton cultivar breeding. Although the nondestructive analysis method of the intact cottonseed can’t substitute for chemical determining or cottonseed kernel powder near-infrared spectroscopy analysis, but it also can be used for screening large numbers of samples. |
