马铃薯主要营养成分的近红外光谱分析

Analysis of Main Nutrients in Potatoes by Near Infrared Spectroscopy

以44个品系马铃薯为原料，利用主成分分析（PCA）方法筛选出代表马铃薯块茎主要营养成分指标（水分、还原糖、淀粉和蛋白质），应用偏最小二乘法（PLS）建立这4种营养成分的预测模型，并对模型预测结果进行了评价。结果表明，马铃薯主要营养成分的模型预测与其相应的化学测量值之间具有较好的相关性，对于水分模型，校正效果：R2cal=98．37％，RMSEE=0．445，RPD=7．84；交叉验证效果：R2cv=93．05％，RMSECV=0．84，RPD=3．79。还原糖模型校正模型效果：R2cal=98．43％，RMSEE=0．0236，RPD=7．99；交叉验证效果：R2cv=86．42％，RMSECV=0．0598，RPD=2．71。淀粉模型校正模型效果：R2cal=97．13％，RMSEE=0．577，RPD=5．9；交叉验证效果：R2cv=95．370％，RMSECV=0．7，RPD=4．65。蛋白质模型校正模型效果：R2cal=98．41％，RMSEE=0．0334，RPD=7．92；交叉验证效果：R2cv=89．49％，RMSECV=0．0767，RPD=3．08。

Potatoes from 44 cultivars were analyzed for some chemical components. The contents of water, reducing sugar, starch and protein were selected as main indicators of the nutrient composition of potatoes using principal component analysis （PCA）. A predictive model for each nutrient indicator was established by partial least square （PLS） method and validated. The established models were found to show good correlation. For the water model, the calibration coefficient of determination （R2ca~）, root mean square error （RMSEE）, and relative predictive determinant （RPD） were 98.38%, 0.445 and 7.84, respectively, and the cross-validation coefficient of determination （R2cv）, root mean square error, and RPD were 93.05%, 0.84 and 3.79, respectively. For the reducing sugar model, the R2caL, RMSEE, and RPD for calibration were 86.42%, 0.0598 and 2.71, respectively, and the R2cv, RMSEE, RPD for cross-validation were 97.13%, 0.577 and 5.9, respectively. For the starch model, the R2c~, RMSEE, and RPD for calibration were 97.13 %, 0.577 and 5.9, respectively, and the R2,, RMSEE, and RPD for cross-validation were 95.370%, 0.7 and 4.65, respectively. For the protein model, the R2c~, RMSEE, and RPD for calibration were 98.41%, 0.0334 and 7.92, respectively, and the R2cv, RMSEE, and RPD for cross-validation were 89.49%, 0.0767 and 3.08, respectively.