Establishment Of Model Of Seeds Protein Content In Peanut Using NIRS And Application In Peanut Breeding

Peanut is one of the main Economic Crops in China,and also is a good source of dietary vegetable protein. The protein content of peanut was studied by Kjeldahl method usually. But this traditional chemical method of these contents have defects, such as high cost, long time consuming, easy to produce pollution, destructive and poor repeatability. Therefore it is necessary to explore a new, fast and nondestructive analysis method of protein content in peanut. In this research, diffuse reflectance spectrums of intact peanut seeds were obtained by near infrared spectrometer, among these spectral data, those of 75 peanuts were analyzed and the model for analysis of protein contents in peanut was established. This model was used to analyze the protein content of other varieties and the hybrids. The main research results were as follows:1. Kjeldahl method was used in this research in order to obtain the protein content of the peanut landraces, and the range of the protein content is from 18.82 % to 25.24 %, and the average is 21.87 %.2. Among these spectral dates, those of 75 peanuts were analyzed and evaluated with modern metrology method. Standard normal variate(SNV),multiplicative scatter correction(MSC),derivative, normalize, smoothing were compared. The results showed that the model developed by means of partial least-squares (PLS) regression,and the spectral data pretreatment of the first derivative + standard normalized variety was the best one. The determination coefficients (R2) of calibration and validation sets were 0.9245 and 0.9018, the standard error of calibration and prediction were 0.3601 and 0.4153, respectively.3.This model was used to predict the protein content of other 98 peanut samples which those samples were studied by traditional chemical methods, compared the two methods. It showed that the protein content of the 98 peanut samples was 18.26 % to 27.08%, and the absolute error was lower 0.5, the relative error less than 2.49%.4. This model was used to analyze the protein content of the 102 hybrid offsprings, it showed that the result is feasible, the high protein varieties could be screening easily.