定量构效关系(QSAR)在环境化学中的进展与方法

定量构效关系(QSAR)是化学品安全评价中有用的工具之一。本文叙述它的应用背景、基本原理、学科历史进展、概念模型、数学模型和算法,着重介绍作者开发的实用量子化学计算程序包,用以产生分子结构参数,所建立的化学物质毒性数据库可用来获取大量生物活性数据,还可用模式识别程序包构筑构效定量关系。本文对其研究前景也作了叙述。     

PCDFs的Ah受体结合能力的定量结构效应关系

应用单苯环氯取代方法取G₁～G₁₀10个分子结构描述符,通过正向逐步线性回归方法构建对于Ah受体结合能力的定量结构-性质相关(QSAR)模型.与有关文献相比,此模型能更好地预测多氯代二苯并呋喃(PCDFs)竞争结合Ah受体logEC₅₀值,各自变量间不存在共线性.模型可以较好地反映PCDFs分子Ah受体结合能力与其分子氯取代位点之间的关系.     

基于QSAR关系预测PCBs和部分PAHs在LDPE中的扩散系数

分子扩散系数（D）是获得污染物与环境介质之间的平衡分配系数（K）的重要前提,然而通过实验测定获取污染物的扩散系数的过程过于繁琐,因此需开发一种更为简单、高效、准确的预测模型来定量预测扩散系数.为此,本文搜集了一些多环芳香烃（PAHs）和多氯联苯（PCBs）在低密度聚乙烯膜（LDPE）上扩散系数（log D）的实测值,基于定量结构-活性关系（QSAR）,利用逐步多元线性回归（MLR）构建了预测D值的模型.模型的决定系数R_adj²为0.941,交叉验证系数Q_LOO²为0.934,外部系数Q_ext²为0.895.结果表明,该QSAR模型具有良好的拟合优度、稳健性和预测能力,其可用来预测应用域内有机污染物在LDPE膜上的扩散系数.     

基于QSAR模型预测有机污染物在XAD与空气中的分配系数

基于定量构效关系(QSAR),运用线性(逐步多元回归MLR)和非线性(支持向量机SVM)两种计算方法开发了两种可靠且高效预测聚苯乙烯二乙烯基苯树脂(XAD)和空气之间分配系数(K_XAD-A)的模型.构建模型的数据包含醇类(Alcohols),苯类(Benzenes),多氯联苯(PCBs)和多环芳香烃(PAHs)等,共计70种有机污染物.两个模型的决定系数R²_adj和外部验证系数Q²_ext均在0.930以上,同时所有物质均在定义的应用域内,结果表明两种QSAR模型有较高的拟合度、稳健性和较为优秀的预测能力,且非线性(SVM)模型比线性(MLR)模型的拟合效果更好.     

苯胺类化合物在不同pH值下对大型蚤的急性毒性及QSAR研究

测定了13种苯胺类化合物在不同pH下（6．0，7．8，9．0）对大型蚤（Daphnia magna)的24h半数活动抑制浓度24h-1C50，应用三种理化参数logP,TSA和pKa，对毒性数据进行了定量构效关系（QSARs)研究，并在此基础上初步探讨了苯胺类化合物的毒性机制。     

取代苯类有机物拓扑指数与酵母菌毒性的人工神经网络

以酿酒酵母菌作为指示生物,对78种取代苯类有机物开展了定量构效关系研究。采用人工神经网络（ANN）建模方法,以酵母菌的1g(1/Cmlz）为活性参数,点价自相关拓扑指数（A、B、C和D值）为分子结构参数,建立了取代苯类有机物定量构效关系神经网络模型;从24种点价自相关拓扑指数中在线筛选出A[0]、A[1]、C[3]、C[5]和D[3]等5种主要结构参数作为ANN的输入节点,并讨论了它们对酵母菌毒性的影响。将模型用于23种取代苯类有机物的生物毒性预测,结果满意。建立了Cmlz与LC50之间线性关系良好的相关性数学模型。     

定量构效关系模型在化合物毒性预测领域的应用、发展与展望

定量构效关系（Quantitative Structure Activity Relationship, QSAR）在化合物毒性虚拟计算领域被广为使用,能够高效、准确地用于化合物毒性的预测和风险评估.本文综述了利用QSAR方法开展化合物毒性预测中数据集、数据输入和模型方法3个领域的发展与变化,分析了研究的发展历程,提出了当前研究热点,并针对定量构效关系在化合物毒性预测领域的未来发展提出了展望.     

部分酯类化合物对四膜虫毒性的全息QSAR分析

全息定量结构活性相关分析(HolographicQSAR或HQSAR)能快速而简便的产生高统计质量和预测能力的QSAR模型。采用HQSAR分析了30个酯类化合物对四膜虫的毒性与分子结构之间的关系,及分子碎片大小和全息长度对QSAR模型的影响,得到的最佳QSAR模型的R2为0.981,Q2为0.912。此外还应用色码讨论了离群值产生的可能原因。     

部分酯类化合物结构与毒性关系研究

该文在化合物二维结构基础上,对48个酯类化合物结构进行了表征,进而建立了化合物结构与水生毒性关系模型,模型的相关系数(R~2)分别为0.926 6和0.917 4,标准偏差(SD)分别为0.257 6和0.255 8。采用留一法交互检验评价了模型的稳定性,得到模型交互检验的相关系数(RCV~2)分别为0.885 6和0.799 2,标准偏差(SD_(CV))分别为0.255 8和0.294 5。运用外部样本检验了模型的预测能力,得到外部预测的相关系数(Rtest2)分别为0.991 0和0.952 3,标准偏差(SD_(test))分别为0.079 3和0.182 9。分子结构描述符能较好地表现化合物结构特征,所建模型拟合效果好、稳定性强、预测准确性高。对于环境中有毒化合物的结构-毒性关系研究具有较好的参考价值。     

支持向量机用于芳烃类化合物对芳烃受体亲和性QSAR研究

尝试将支持向量机(SVM)应用于3种典型芳烃类环境毒物(PCDD,PCDF和PCB)定量构效关系研究,通过对芳烃受体亲和性考察,结果发现该组样本的生物活性在一定程度上与分子电性距离矢量具有非线性联系.SVM对内部和外部样本都具良好稳定性能和预测能力:所得模型拟合、交叉检验、外部预测复相关系数及均方根误差分别为R2cum=0.922、Q2cum=0.825、Q2ext=0.834和RMSext=0.531将其与文献报道及多元线性回归、偏最小二乘、人工神经网络进行比较,结果表明对小样本、非线性问题SVM具较强拓展性及泛化能力,故在环境毒物评价和控制中具有广阔应用前景.     

The impact of screening for open neural tube defects in England and Wales

Using information derived from the voluntary system of notification of congenital malformations in England and Wales, the birth prevalence of anencephaly and spina bifida was estimated to have declined by 80 per cent from 31.5 to 6.2 per 10 000 between 1964–1972 and 1985. Over the same period, notified terminations of pregnancy with a suspected fetal central nervous system abnormality increased from less than 1 per cent to 56 per cent of neural tube defect births and central nervous system terminations combined, accounting for 31 per cent of the decline in births. Routinely collected national statistics provide a method for monitoring the impact of screening for open neural tube defects. However because they are incomplete and lack detail an alternative method of monitoring is needed. This paper includes an outline of such a method, together with the results of a pilot study designed to assess the feasibility of monitoring screening in the Oxford Region.     

利用人工神经网络对空气中O3浓度进行预测

将人工神经网络应用于对空气中O3的浓度预测,提出了完整的预测模型,选取风速、风向、相对湿度、云量、平均气温、最高气温等6项气象因子作为输入量,经过两个月的预测实验,结果表明,实测值与预测值的平均相对误差为21.49%,相关系数为0.837.表明人工神经网络对O3的浓度预测是一种有效的工具.     

人工神经网络模型在石油资源预测中的应用

锦270井区大凌河油层是盆地大幅度沉降条件下的产物，以深陷湖环境为主，发育湖底扇浊积岩。准确预测大凌河油层的砂体分布和泥质含量，是勘探目标选择的关键。由于钻井资料少，不能反映储层横向变化。为了准确预测储层的横向变化，综合钻井地质资料和地震勘探资料，并为了反映地震资料的多个参数与储层横向变化间的非线性相关关系，采用人工神经网络模型进行预测，计算了大凌河油层砂层厚度和泥质含量的平面分布。依据计算结果，分析了有利油气区的分布，并给出了几点结论。     

Prenatal screening and diagnosis of neural tube defects in England and Wales in 1985

A survey was carried out to determine the effect of prenatal screening and therapeutic abortion on births in 1985 with anencephaly and spina bifida in England and Wales. Maternal serum alpha-fetoprotein tests were done on 399 288 women (60 per cent of pregnant women): 4 per cent were reported as being screen-positive and 1 per cent had an amniocentesis. An estimated 534 pregnancies associated with anencephaly were terminated and an estimated 445 pregnancies associated with spina bifida (but without anencephaly) were terminated. Most (63 per cent) of the anencephalic pregnancies were first suspected from an ultrasound examination; 57 per cent of the spina bifida pregnancies were first suspected from a positive maternal serum alpha-fetoprotein test, 35 per cent by ultrasound, and the remaining 8 per cent by other means. The birth prevalence of anencephaly declined by 94 per cent between 1964–1972 and 1985, but when the terminations of pregnancy on account of having a fetus with anencephaly are added to the births the decline in prevalence was only 50 per cent. The birth prevalence of spina bifida declined by 68 per cent over the same period but when the terminations were added to the births the decline in prevalence was only 32 per cent. Among births with anencephaly 66 per cent had had no screening or diagnostic tests in early pregnancy, but in those that did nearly all were positive–usually in twin pregnancies where one fetus was affected but not the other. Among births with spina bifida, 48 per cent had no tests and in those that did the results were mainly negative. We conclude that in order to monitor adequately the national screening programme for anencephaly and spina bifida a special neural tube defects register should be formed.     

Enhanced twin pregnancy detection within an open neural tube defect and down syndrome screening protocol using free-beta hCG and AFP

We have applied our multimarker approach of maternal serum alpha-fetoprotein (AFP) and free-beta human chorionic gonadotropin (hCG) for Down syndrome screening to multiple gestations to assess its efficacy for improved detection of twin and triplet pregnancies. This study matched 225 cases of twin pregnancy and 39 cases of triplet pregnancy each with ten singleton pregnancies based on gestational week, race, time to receive sample, time of year of sample, and geographical area. The ratios of the MOM for each group at the tenth, 50th, and 90th percentiles were compared by the Wilcoxon test. Risks for twins were calculated using Bayes' rule, the age-related incidence of twins, and the levels of AFP and free-beta hCG. The tenth, 50th, and 90th percentiles of free-beta hCG MOMs in twin and triplet cases were 0.85, 1.99, and 4.51, and 1.38, 2.78, and 4.07, respectively. For AFP, the MOMs at these percentiles were 1.26, 1.91, and 2.99, and 2.02, 2.68, and 5.30, respectively. The twin and triplet distributions for each marker were statistically significantly different from the singleton distributions (P<0.0001) and from each other (P=0.0012). At a twin risk cut-off of 1 in 50, 77.4 per cent of all twin gestations can be detected in a second-trimester AFP and free-beta hCG screening protocol with 5.1 per cent of singleton pregnancies falsely identified as at risk for twins. Our dual marker protocol for mid-trimester pregnancy screening combining AFP and free-beta hCG can identify over 77 per cent of twin pregnancies in women less than 35 years of age. This benefit may contribute to an improved outcome of pregnancy by early detection of multiple gestation.     

基于模糊识别和遗传神经网络的渤海湾叶绿素a预测研究

由于海洋生态系统的高度复杂性和非线性,利用新兴的水信息学技术,包括模糊模式识别、遗传算法、人工神经网络等构建渤海湾叶绿素a预测模型.以渤海湾实测水质数据为依据,利用遗传算法的全局搜索能力优化BP神经网络的初始权值和阈值,从而避免陷入局部最优解,构建GANN叶绿素a预测模型.模型预测的均方根误差为3.81 μg/L,仿真...     

The impact of maternal serum alpha fetoprotein screening on open neural tube defect births in North-East Scotland

Over the three years period 1980–1982, 18 256 pregnancies in the Grampian Region of N-E Scotland including the islands of Orkney and Shetland were screened for raised levels of maternal serum alpha fetoprotein (MSAFP) in the second trimester. Thirty six cases of fetal open neural tube defect in singletons were detected (18 anencephaly and 18 spina bifida). Four additional cases of open spina bifida were associated with normal MSAFP levels although two of these were detected by amniotic fluid AFP measurement when amniocentesis was carried out because of previous NTD history. A further three cases of open spina bifida and two of anencephaly occurred in unscreened pregnancies. The MSAFP screening programme alone was thus instrumental in reducing the birth incidence of open neural tube defects by 36 out of 45 cases (80 per cent) in singletons.     

Integration of artificial neural network and MADA methods for green supplier selection

In recent years, with the raising of awareness in environmental protection and sustainable development in enterprises, the green issue has become more and more critical in supply chain management. This study intends to develop a green supplier selection model which integrates artificial neural network (ANN) and two multi-attribute decision analysis (MADA) methods: data envelopment analysis (DEA) and analytic network process (ANP). It is called ANN–MADA hybrid method. ANN–MADA hybrid method considers both practicality in traditional supplier selection criteria and environmental regulations. It also overcomes traditional DEA drawbacks, limitations of data accuracy and decision-making units (DMUs) amounts constraint. The model evaluation results of an international well-known camera manufacturer indicate that the ANN–MADA hybrid method outperforms two other hybrid methods, ANN–DEA and ANP–DEA. It was also discovered that ANN–MADA has better power of discrimination and noise-insensitivity in evaluating green suppliers’ performances.     

自组织神经网络在有害赤潮预警研究中的应用

根据非线性理论及有害赤潮的特点，探讨应用人工神经网络原理进行赤潮预警预报的方法，提出网络可塑性较强，无监督的ART－KOHONEN网络模型，阐述了其基本原理和算法。将该模型应用于辽宁省营口市鲅鱼圈海域，结果表明：该方法能较理想地解决无训练样本的分类识别问题，具有较高的精度，可用于有害赤潮的预警预报工作。     

Multiple linear regression models of urban runoff pollutant load and event mean concentration considering rainfall variables

Rainfall is an important factor in estimating the event mean concentration (EMC) which is used to quantify the washed-off pollutant concentrations from non-point sources (NPSs). Pollutant loads could also be calculated using rainfall, catchment area and runoff coefficient. In this study, runoff quantity and quality data gathered from a 28-month monitoring conducted on the road and parking lot sites in Korea were evaluated using multiple linear regression (MLR) to develop equations for estimating pollutant loads and EMCs as a function of rainfall variables. The results revealed that total event rainfall and average rainfall intensity are possible predictors of pollutant loads. Overall, the models are indicators of the high uncertainties of NPSs; perhaps estimation of EMCs and loads could be accurately obtained by means of water quality sampling or a long term monitoring is needed to gather more data that can be used for the development of estimation models.