Combination of near Infrared Spectroscopy and Multivariate Analysis to Detect Valine (c5h11no2 ) in Oilseed Rape Leaves

Valine (C5H11NO2) is one of the most important amino acids in oilseed rape and has a close relationship with the influence of herbicide stress during growing stage. The traditional methods for amino acid detection were to use high performance liquid chromatography (HPLC) or amino acid analyzer. These methods were time consuming for preparation and detection, costly, laborious and not convenient for fast determination during each growing status of oilseed rape. Therefore, near infrared (NIR) spectroscopy (1100-2500 nm) combined with multivariate calibrations was investigated to determine valine in oilseed rape leaves under herbicide stress. The calibration set consisted of 186 leave samples, and the validation set consisted of 62 samples. Different preprocessing methods were compared, including multiplicative scatter correction (MSC), first-derivative (1-Der), second-derivative (2-Der), de-trending and direct orthogonal signal correction (DOSC). The optimal performance was determined by the PLS model with correlation coefficients (r) and root mean squares error of prediction (RMSEP). The optimal PLS model was achieved by DOSC spectra with r=0.9822 and RMSEP=0.0379. Successive projections algorithm (SPA) was recommended for variable selection procedure. After computation, the number of selected variables was 15, 17 and 1 for 1-Der, de-trending and DOSC spectra, respectively. The optimal prediction results was achieved by SPA-LS-SVM model (De-trending spectra) with r=0.9864 and RMSEP=0.0327. This result was slightly better than PLS model with DOSC spectra. The results indicated that near infrared spectroscopy combined with SPA-LS-SVM was successfully applied for the determination of valine in oilseed rape leaves under herbicide stress.