The Department of Medical Nursing, Hulunbuir Vocational Technical College, Hulunbuir 021000, China.
Department of Pathology, Hulunbeir People's Hospital, Hulunbuir 021008, China.
Comput Intell Neurosci. 2022 Sep 1;2022:9022821. doi: 10.1155/2022/9022821. eCollection 2022.
This work was to explore the application value of gastrointestinal tumor markers based on gene feature selection model of principal component analysis (PCA) algorithm and multicolor quantum dots (QDs) immunobiosensor in the detection of gastrointestinal tumors. Based on the PCA method, the neighborhood rough set algorithm was introduced to improve it, and the tumor gene feature selection model (OPCA) was established to analyze its classification accuracy and accuracy. Four kinds of coupled biosensors were fabricated based on QDs, namely, 525 nm Cd Se/Zn S QDs-carbohydrate antigen 125 (QDs525-CA125 McAb), 605 nm Cd Se/Zn S QDs-cancer antigen 19-9 (QDs605-CA19-9 McAb), 645 nm Cd Se/Zn S QDs-anticancer embryonic antigen (QDs 645-CEA McAb), and 565 nm Cd Se/Zn S QDs-anti-alpha-fetoprotein (QDs565-AFP McAb). The quantum dot-antibody conjugates were identified and quantified by fluorescence spectroscopy and ultraviolet absorption spectroscopy. The results showed that the classification precision of OPCA model in colon tumor and gastric cancer datasets was 99.52% and 99.03%, respectively, and the classification accuracy was 94.86% and 94.2%, respectively, which were significantly higher than those of other algorithms. The fluorescence values of AFP McAb, CEA McAb, CA19-9 McAb, and CA125 McAb reached the maximum when the conjugation concentrations were 25 g/mL, 20 g/mL, 30 g/mL, and 30 g/m, respectively. The highest recovery rate of AFP was 98.51%, and its fluorescence intensity was 35.78 ± 2.99, which was significantly higher than that of other antigens ( < 0.001). In summary, the OPCA model based on PCA algorithm can obtain fewer feature gene sets and improve the accuracy of sample classification. Intelligent immunobiosensors based on machine learning algorithms and QDs have potential application value in gastrointestinal gene feature selection and tumor marker detection, which provides a new idea for clinical diagnosis of gastrointestinal tumors.
本研究旨在探讨基于主成分分析（PCA）算法和多色量子点（QDs）免疫生物传感器的胃肠肿瘤标志物基因特征选择模型在胃肠肿瘤检测中的应用价值。基于 PCA 方法，引入邻域粗糙集算法对其进行改进，建立肿瘤基因特征选择模型（OPCA），分析其分类精度和准确率。基于量子点构建了 4 种偶联型免疫生物传感器，即 525nm CdSe/ZnS QDs-糖类抗原 125（QDs525-CA125 McAb）、605nm CdSe/ZnS QDs-癌抗原 19-9（QDs605-CA19-9 McAb）、645nm CdSe/ZnS QDs-癌胚抗原（QDs645-CEA McAb）和 565nm CdSe/ZnS QDs-甲胎蛋白（QDs565-AFP McAb）。通过荧光光谱和紫外吸收光谱对量子点-抗体偶联物进行鉴定和定量分析。结果表明，在结肠肿瘤和胃癌数据集上，OPCA 模型的分类精度分别为 99.52%和 99.03%，分类准确率分别为 94.86%和 94.2%，均显著高于其他算法。AFP McAb、CEA McAb、CA19-9 McAb 和 CA125 McAb 的荧光值在偶联浓度分别为 25μg/mL、20μg/mL、30μg/mL 和 30μg/mL 时达到最大值。AFP 的最高回收率为 98.51%，其荧光强度为 35.78±2.99，显著高于其他抗原（<0.001）。综上所述，基于 PCA 算法的 OPCA 模型可以获得更少的特征基因集，提高样本分类的准确性。基于机器学习算法和量子点的智能免疫生物传感器在胃肠肿瘤基因特征选择和肿瘤标志物检测方面具有潜在的应用价值，为胃肠道肿瘤的临床诊断提供了新的思路。
