| The objective of the present study was to collect corn and wheat samples from different planting regions and varieties from all over china, respectively, and evaluate their nutritional values and variation. Prediction equations were developed for the rapid determination of proximate nutrients and amino acids in corn and wheat by near-infrared reflectance spectroscopy (NIRS). In addition, according to the reference data (digestible energy and metabolizable energy values) determined by digestion-metabolism experiments using growing pigs, the possibility of using NIRS for the rapid determination of available energy (digestible energy, DE; metabolizable energy, ME) in corn and wheat was also investigated. This thesis includes the following4experiments:Experiment1:The possibility of using NIRS for quantitative determination of proximate nutrients and swine available energy in corn was investigated. From2009to2011, a total of117corn samples were collected from all over China. The proximate composition were analyzed, and the DE and ME of these corn samples were determined by digestion-metabolism experiments using growing pigs. Results showed that the variations in nutrients were large between different sources of the corn samples. The coefficient of determination for calibration (RSQcal), coefficient of determination for cross-validation (1-VR), ratio of performance to deviation for cross-validation (RPDcv), coefficient of determination for external validation (RSQv) and ratio of performance to deviation for validation (RPDv) of moisture, crude protein (CP), ether extract (EE), acid detergent fiber (ADF), neutral detergent fiber (NDF) and density was0.89-0.95,0.87-0.93,2.83-3.67(>2.50),0.85-0.91and2.67-3.20(>2.50), respectively, which indicates that these NIRS models can be used in routine analysis. The predictive performance of ash, total phosphorus (TP) and starch was poor (RPDv,1.92-2.47,<2.50) and could not be used for routine analysis. The RSQca,,1-VR, RPDcv, RSQv and RPDv for GE, DE and ME was0.87-0.94,0.86-0.92,2.78-3.53(>2.50),0.86-0.90and2.64-3.17(>2.50), respectively, which indicates that good NIRS models were obtained for these three energy constituents and these prediction equations can be used in routine analysis.Experiment2:The possibility of using NIRS for quantitative determination of amino acids of corn was investigated. From2009to2011, a total of89corn samples were collected from all over China, and the18amino acids contents were analyzed. Results showed that the variations in18amino acids were large between different sources of the collected corn samples. Except for the lysine, methionine, tryptophan and cystine, good NIRS prediction equations were obtained for the other14amino acids with high RSQcal (0.86-0.94),1-VR (0.84-0.93) and RPDv (2.56-4.44,>2.50) values. Excellent predictive performance of the NIRS models were received for most amino acids (RSQv,0.83-0.91; RPDv,2.51-3.33,>2.50) with the exception of arginine, lysine, methionine, tryptophan, cysteine, glycine and tyrosine, and these NIRS equations could be used in routine analysis. Except for tryptophan, compared to the linear regression results of amino acid contents relative to crude protein, NIRS has a better predictive performance. Experiment3:The possibility of using NIRS for quantitative determination of proximate nutrients and swine available energy in wheat was investigated. From2011to2012, a total of58wheat samples were collected from all over China. The proximate composition were analyzed, and the DE and ME were determined by digestion-metabolism experiments using growing pigs. Results showed that the variations in nutrients were large between different sources of the wheat samples. The RSQcal,1-VR, RPDcv, RSQv and RPDv of the NIRS models developed by44calibration samples for moisture, CP, EE, NDF, density and GE was0.87-0.92,0.85-0.90,2.68-3.13(>2.50),0.84-0.90and2.51-3.16(>2.50), respectively, which indicates that these prediction equations can be used in routine analysis. The predictive performance of ADF, ash, TP and starch was poor (RPDv,1.91-2.43,<2.50) and could not be used for routine analysis. The NIRS models obtained for DE and ME also can not be used for routine analysis with low RPDcv (2.29-2.41,<2.50) and RPDv (2.03-2.11,<2.50). The cross-validation performance of models developed by58calibration samples was improved, and the RPDcv values exceeded2.50with the exception of ash, TP, starch and ME.Experiment4:The possibility of using NIRS for quantitative determination of amino acids of wheat was investigated. From2009to2011, a total of450wheat samples were collected from all over China,381samples were chosen for NIRS calibration and validation by CENTER and SELECT algorithm, and the18amino acids contents were analyzed. Results showed that the variations in18amino acids were large between different sources of the wheat samples. Except for methionine, valine, tryptophan, cysteine and tyrosine, the RSQcal,1-VR, RPDcv, RSQv and RPDv for the other13amino acids was0.87-0.96,0.83-0.95,2.53-4.53(>2.50),0.83-0.9and2.54-3.93(>2.50), respectively, which indicates that these NIRS models can be used in routine analysis. Except for tryptophan, the predictive performance measuring amino acids by NIRS was better than that obtained by crude protein regression. |
PDF Full Text Request