摘要
环境中的微/纳米塑料污染引起了人们极大关注。土壤中的微/纳米塑料不可避免对植物产生影响,因此预测微/纳米塑料的植物毒性可为土壤中微/纳米塑料治理提供抓手。以水稻为研究对象,发展了基于同步辐射X射线荧光(SRXRF)光谱与机器学习的非靶标金属组学方法,以预测聚氯乙烯纳米塑料(nPVC)对水稻的毒性。首先将水稻暴露于不同浓度(500 ng/g与500μg/g)nPVC中,培养35 d后,收集水稻叶;其次,利用SRXRF研究暴露nPVC后水稻叶中金属组的变化;然后,利用机器学习方法区分暴露不同浓度nPVC水稻样品。对SRXRF光谱进行主成分分析(PCA)非监督聚类,发现500μg/g组能够良好聚类,而500 ng/g组与对照组无明显差异,表明500 ng/g的nPVC暴露对植物的毒性远低于500μg/g nPVC。对SRXRF全光谱,利用线性模型k近邻算法(kNN)和非线性模型支持向量机(SVM)建立预测模型,区分不同组别的准确率可达94.12%。为了提升运算速度,减少模型计算量,使用竞争性自适应加权重采样算法(CARS)挑选特征光谱建立预测模型,区分不同组别的准确率为89.51%。相对全光谱模型,特征光谱预测模型虽然预测准确率下降了4.61%,但模型输入参数减少了99.38%,因此同样具有良好潜力。研究表明,基于SRXRF和机器学习的非靶标金属组学可准确预测不同浓度nPVC对水稻金属组的干扰程度,从而反映nPVC对水稻毒性的浓度依赖性。方法同样可用于预测其他微/纳米塑料毒性的浓度依赖性。
Micro/nano plastics pollution in the environment has attracted great attention.Micro/nano plastics in soil inevitably pose a threat to the plants,so predicting the phytotoxicity of micro/nano plastics can provide a lever for the treatment of micro/nano plastics in soil.In this paper,a non-target metallomics method based on synchrotron radiation X-ray fluorescence(SRXRF)spectra and machine learning was developed to predict the toxicity of polyvinyl chloride nanoplastics(nPVC)to rice.Firstly,rice was exposed to different concentrations of nPVC(500 ng/g and 500μg/g),and rice leaves were collected after 35 days of cultivation.Secondly,SRXRF was used to study the changes of metal groups in rice leaves after exposure to nPVC.The unsupervised clustering of principal component analysis(PCA)on SRXRF spectra showed that the 500μg/g group was able to cluster well,while there was no significant difference between the 500 ng/g group and the control group,indicating that the toxicity of nPVC exposure of 500 ng/g to plants was much lower than that of 500μg/g nPVC.For the full-channel spectra of SRXRF,the linear model K-nearest neighbor(KNN)and nonlinear model support vector machine(SVM)algorithm were established,and the accuracy of different groups was 94.12%.The competitive adaptive reweighted sampling(CARS)algorithm was used to select the characteristic spectra to establish the prediction model,and the accuracy of distinguishing different groups was 89.51%.Compared with the full-channel spectral model,although the accuracy of the characteristic channel model is reduced by 4.61%,the input parameters of the model are reduced by 99.38%,so the excellently potential of the model was showcasing.This study showed that non-target metallomics based on SRXRF and machine learning could accurately predict the interference degree of different concentrations of nPVC on the metal group of rice,so as to reflect the concentration dependence of nPVC on rice toxicity.This method can also be used to predict the concentration-dependent toxicit
作者
魏超杰
解宏鑫
王伟
李柏
李玉锋
WEI Chaojie;XIE Hongxin;WANG Wei;LI Bai;LI Yu-Feng(College of Engineering,China Agricultural University,Beijing 100083,China;CAS-HKU Joint Laboratory of Metallomics on Health and Environment,&CAS Key Laboratory for Biomedical Effects of Nanomaterials and Nanosafety,&Beijing Metallomics Facility,&National Consortium for Excellence in Metallomics,Institute of High Energy Physics,Chinese Academy of Sciences,Beijing 100049,China)
出处
《中国无机分析化学》
CAS
北大核心
2024年第8期1015-1021,共7页
Chinese Journal of Inorganic Analytical Chemistry
基金
国家自然科学基金面上项目(32272410)。
关键词
聚氯乙烯纳米塑料
水稻
同步辐射X射线荧光光谱
机器学习
非靶标金属组学
polyvinyl chloride nanoplastics
rice plant
synchrotron radiation X-ray fluorescence
machine learning
non-targeted metallomics
