當人類及動物攝取含有毒素汙染的穀物時,會產生不同程度的中毒徵兆。人類長期食入低汙染劑量的穀物,會產生免疫抑制及生長不佳的狀況,且有致癌之可能性,而動物在短時間內食入大量毒素則可能導致死亡。本實驗嘗試開發以近紅外線光譜儀技術,檢測飼料穀物毒素中的伏馬鐮孢毒素FB1含量。運用線性迴歸(SLR)與部分最小平方迴歸(PLS)法建立NIR光譜與抗原抗體免疫親和分析值之間的多元校正模型,並對21件穀物樣品進行含量預測。分析結果以線性迴歸建立之檢量線之預測能力較佳,標準測定誤差(SEP)為4.05;若將線性迴歸檢量線單獨套用在玉米樣品,標準測定誤差(SEP)可降低為0.05。而部分最小平方以一次微分、波數範圍5334.14cm-1-5260.86cm-1,PLS Factor為4時,有最佳的預測能力,標準測定誤差(SEP)為1.760,遠不如線性迴歸。建議以增加樣品數與持續尋找最佳化條件,藉由部分最小平方迴歸可降低雜訊影響的特性,建立良好的檢測模式。
Ingestion of mycotoxin contaminated cereal foods will lead to various poisoned symptom in animals and human. For long term ingestion of low level contamination, the human may develop symptoms of immunity depression, even carcinogen accumulation. In animals, high dose of mycotoxins in short time may be lethal. This research studied the quantification of Fumonisins by Near Infra Red (NIR) spectrum technology. The NIR absorption spectrum were analyzed by Linear Regression (SLR), and Partial Least Square (PLS) methodologies, to construct a multivariable correction model for establishing correlation between data from NIR and the result of using immuno-affinity column. Twenty one cereal samples were used for this purpose. It was found that SLR methodology provided better prediction, and the standard error (SEP) was 4.05. For SEP application on corn samples the SEP was reduced to 0.05. For PLS, in the wavelength range of 5334.14 cm-1 to 5260.86 cm-1, the best prediction capability was obtained when PLS factor was set at 4. The SEP was 1.760, which was inferior to that of SLR. Increasing the number of samples and fine tuning the optimal operation condition, applying PLS may help reducing the interfering effects of inherent noise and build a prediction model for quantifying mycotoxin concentrations.