葡萄牙牡蛎( Crassostrea angulata)是中国南方重要的海水养殖生物。 由于长期的人工育苗, 且未采取种质保护措施, 其种质资源不断衰退, 从而给牡蛎消费市场带来了不良影响, 急需开展葡萄牙牡蛎良种选育工作。 选育营养好、 口感佳的葡萄牙牡蛎, 需要对大量样本进行营养成分分析, 传统的实验室化学法耗时费力, 成本较高, 因此需寻求一种高效的葡萄牙牡蛎化学成分含量测定方法。 采用美国Thermo Fisher公司的傅里叶近红外光谱仪对6个产地的105份经过冷冻干燥和研磨处理的葡萄牙牡蛎样本(去除闭壳肌)进行光谱扫描, 通过采集的光谱数据与测定的化学真实值对比分析, 检测了近红外光谱技术(NIRS)对葡萄牙牡蛎中蛋白质、 糖原、 牛磺酸、 锌、 硒和钙6种成分含量预测的准确性。 利用TQ Analyst软件, 选用偏最小二乘法(PLS), 乘法散射校正(MSC)、 一阶求导、 Norris平滑等光谱预处理方法, 建立了6种成分的近红外定量模型, 并选取1/3总样品量的样本作为验证样本, 对模型进行了外部验证和交叉验证。 葡萄牙牡蛎的蛋白质、 糖原、 牛磺酸、 锌、 硒和钙6个模型的校正相关系数( R_C)分别为0.985 3, 0.965 1, 0.950 4, 0.955 4, 0.920 0和0.925 2, 预测相关系数( R_P)分别为0.985 1, 0.963 6, 0.944 1, 0.946 1, 0.919 0和0.924 1, 交叉验证相关系数( R_CV)分别为0.981 7, 0.946 1, 0.900 5, 0.897 5, 0.875 3和0.829 2。 结果表明, 模型预测值与化学真实值有很高的相关度, 近红外光谱技术可以比较准确地预测葡萄牙牡蛎中蛋白质、 糖原、 牛磺酸、 锌、 硒、 钙的含量。 本实验样本采集时间跨度长, 产地分布范围广, 数量大, 具有较好的代表性, 样本经过冷冻干燥处理, 减少了水分对光谱质量的影响, 提高了模型的准确性及稳定性。 鉴于近红外光谱技术分析过程高效, 不使用化学试剂, 检测成本低, 该模型的建立对开展大规模葡萄牙牡蛎营养成分快速分析, 选育肉质性状佳的新品系具有重要意义。

Crassostrea angulata is the main variety of marine aquaculture in southern China. Due to long-term artificial breeding with no germplasm protection measures, its germplasm resources are declining, which has a negative impact on the oyster consumption market. Therefore, it is urgent to develop breeding of Crassostrea angulata ( C. angulata). The selection for C. angulata with good nutrition and good flesh quality requires a large number of samples in the nutrient analysis. Traditional laboratory chemical method is time-consuming and costly, so we are looking for an efficient method for determining the chemical content of C. angulata. The spectroscopic scan was carried out using 105 frozen-dried and grinded C. angulata samples (removed the adductor muscle) from six regions with the Fouriernear-infrared spectrometer (Thermo Fisher, USA) in this study. By comparing the spectroscopic scan data to the chemical values, the accuracy of the content predictions of protein, glycogen, taurine, zinc, selenium and calcium in C. angulata obtained by near-infrared spectroscopy (NIRS) was studied. Using TQ Analyst (Thermo Fisher, USA) software, and selecting partial least squares (PLS), spectral preprocessing method like multiplication scattering correction (MSC), 1st derivative, and Norris derivative filter, the near-infrared models of the six components were established. And 1/3 of the total samples were selected as validation samples. The models were validated by external and cross validation. The correlation coefficients of calibration ( R_C) of the six models of protein, glycogen, taurine, zinc, selenium and calcium were 0.985 3, 0.965 1, 0.950 4, 0.955 4, 0.920 0 and 0.925 2, respectively. The correlation coefficients of prediction ( R_P) were 0.985 1, 0.963 6, 0.944 1, 0.946 1, 0.919 0 and 0.924 1, respectively. The correlation coefficients of cross validation ( R_CV) were 0.981 7, 0.946 1, 0.900 5, 0.897 5, 0.875 3 and 0.829 2, respectively. The results showed that the predicted values of the models had a high correlation with the chemical values, which indicated the NIRS could accurately predict the contents of protein, glycogen, taurine, zinc, selenium and calcium in C. angulata. The samples in this study had good representativeness. The collection time was long. The production area was wide and the quantity was large. The samples were frozen-dried, which reduced the influence of water on the spectral quality. Thus, the accuracy and stability of the models were improved. Spectroscopic scan and analysis based on NIRS was very efficient with no chemical reagents and low cost. The established quantitative model for 6 chemical compositions in C. angulata by NIRS would have a great significance for large-scale analysis of the nutritional compositions and for the selection of new strains with good flesh quality in C. angulata.