基于岭回归和SVM的高维特征选择与肽QSAR建模

岭回归估计权重绝对值在一定程度上体现了对应特征作用大小, 据此发展了基于岭回归(RR)和支持向量机(SVM)的高维特征选择算法. 对苦味二肽(BTT)和细胞毒性T淋巴细胞(CTL)表位9 肽两个肽体系, 以氨基酸的531 个物理化学性质参数直接表征肽结构, 各获得1062、4779 个初始特征; 对训练集, 初始特征以岭回归排序后序贯引入, 当SVM留一法交叉测试(LOOCV)的均方误差(MSE)显著上扬时终止, 最后以多轮末尾淘汰进一步精筛, 分别获得7、18个物理化学意义明确的保留特征. 基于保留特征与支持向量回归(SVR), 对训练集建立定量构效关系(QSAR)模型, 预测独立测试集, 其拟合精度、留一法交叉测试精度、独立预测精度均优于现有文献报道结果. 新方法运行速度快, 选取的特征物理化学意义明确, 解释性强, 在肽、蛋白质定量构效关系建模等高维数据回归预测领域有较广泛应用前景.     

4种津类除草剂的气相色谱-电子轰击质谱裂解规律研究

根据4种均三嗪类除草剂(特丁津、另丁津、草达津和扑灭津)在EI源离子化条件下所得到的质谱图,对这4种化合物裂解生成的特征离子及其裂解方式和机理作了比较研究。发现此类化合物虽具有相同母体和近似的结构,但所具有的微细结构差异引起质谱图的明显不同。利用此裂解特征可对同类化合物的衍生物或异构体作出鉴别和确认。     

超高效液相色谱-四极杆/静电场轨道阱高分辨质谱法非靶向筛查苹果中苯脲类农药

建立了超高效液相色谱－四极杆/静电场轨道阱高分辨质谱（UPLC-Q/Orbitrap HRMS）非靶向筛查苹果中苯脲类农药的方法。样品采用QuEChERS法提取净化,Acquity BEH C18色谱柱（100 mm × 2.1 mm,1.7 μm）分离,以甲醇和含0.1%甲酸的水溶液为流动相进行梯度洗脱,在电喷雾正离子模式下采用四极杆/静电场轨道阱高分辨质谱进行检测。将13种苯脲类除草剂和9种苯甲酰脲类杀虫剂按化学结构分为4类。首先通过对4类22种典型苯脲类农药标准品的准分子离子和二级质谱碎片进行分析,总结苯脲类农药的质谱裂解规律如下：绿麦隆等9种苯脲类除草剂的主要特征离子碎片为m/z 72.044 59,可通过特征丢失中性分子二甲胺（m/z 45.058 03）产生特征离子碎片;绿谷隆等4种苯脲类除草剂可通过特征丢失中性分子甲醇［CH3OH］或卤化氢［HR1］（R=Cl,Br,F）产生离子碎片;除虫脲等7种含氟苯甲酰脲杀虫剂的主要特征离子碎片为 m/z 158.040 47、141.015 00,也可发生特征中性丢失2,6-氟苯甲酰胺结构［C8H3F2O2NH2］（m/z 183.013 21）;杀铃脲等2种含氯苯甲酰脲类杀虫剂的主要特征离子碎片为m/z 156.020 25、138.993 76、113.015 28。利用该方法对北京12份市售苹果进行非靶向筛查,在1份样品中筛查出绿麦隆。该方法可为快速筛查农产品中相似结构特征的苯脲类化合物提供参考。     

土壤中多种有机磷及氨基甲酸酯类农药残留量的气相色谱-质谱法测定

建立了气相色谱-质谱法测定土壤中12种有机磷和氨基甲酸酯类农药残留分析方法。以丙酮-石油醚(4∶1,V/V)为提取剂,采用超声波提取土壤中农药残留,经弗罗里土层析柱净化,气相色谱-质谱(选择离子模式)法同时测定了土壤中多种有机磷和氨基甲酸酯类农药。该法对0.1μg/mL和0.5μg/mL两个浓度添加水平的回收率分别为70.1%~119.0%和78.1%~119.1%,相对标准偏差分别为6.30%~9.80%和5.20%~8.23%。     

固相萃取-气相色谱-质谱联用同时测定河水和海水中87种农药

采用固相萃取-气相色谱-质谱联用技术,建立了河水和海水中87种农药(24种有机磷、15种有机氯、12种唑类、9种拟除虫菊酯类、5种氨基甲酸酯类、7种酰胺类及15种其他新型农药)的多残留同时分析方法。优化了影响分离效果和灵敏度的仪器参数,考察了固相萃取柱柱型及水样体积、pH、盐度的影响,采用NH2柱优化了净化效果,内标法和替代物法用于数据的质量控制。结果表明: 在最佳条件下,各目标农药的方法检出限为0.1～6.6 ng/L;以实际河水和海水为基底,在5 ng/L和20 ng/L的加标水平下,绝大多数目标农药的回收率为60%～120%,相对标准偏差(n=4)为0.01%～9.7%。该法灵敏、准确,已成功地应用于福建九龙江河口区表层水样中多种类农药的复合污染监测,检出包括5种有机磷类、3种酰胺类、4种唑类、3种氨基甲酸酯类、2种拟除虫菊酯类等农药20种。     

机器学习方法用于激素敏感脂肪酶抑制剂活性预测

对激素敏感脂肪酶,我们构建了表征分子组成、电荷、拓扑、几何结构及物理化学性质等特征的1559个描述符,通过Fischer Score排序过滤和Monte Carlo模拟退火法相结合进行变量筛选得到35个描述符,然后分别用支持向量学习机(SVM)、人工神经网络(ANN),k-近邻(k-NN),连续核密度估计(CKD)和逻...     

基于Person Ⅶ核函数的支持向量机及其在化学模式分类中的应用

化学模式分类问题通常是非线性的,而且比较复杂,难以用经典统计方法建立分类判别模型。以支持向量机(SVM)构建的分类器具有更好的分类性能。对于非线性分类,SVM通过核函数将其映射到高维特征空间中,然后再进行线性分类。因此,核函数往往是决定SVM非线性分类性能的关键。实际应用时,一般通过选择几种核函数,并对其参数进行优化,然后根据分类器的预测性能来决定,训练过程非常耗时,而且结果难以保证最优。为此,采用一种通用性的核函数,即PersonⅦ核函数(PUKF),它可取代目前常用的几种核函数,可避免SVM非线性分类器训练过程的核函数选择问题。本研究将基于PUKF的SVM分类器应用于两个化学模式分类问题,均取得了较好的结果。对于多类分类,设计了一种子分类器的构造方法,它在分类性能保持较好的情况下,简化了多类分类器结构,大大降低了计算量。     

紫杉烷二萜类化合物的定量构效关系研究

紫杉醇是从紫杉或红豆杉树中提取的一种天然抗癌原料药,具有独特的抗癌机理。由于紫杉醇的种种限制,开发具有更高抗癌活性的类紫杉醇药物具有广阔的前景。紫杉烷二萜是以紫杉醇为母体,通过对其结构的不断修饰得到的一些二代紫杉醇类化合物。本文选用30个结构多样的紫杉烷二帖类化合物作为数据集,随机选取其中24个作为训练集,其它分子作为检验集,采用多元线性回归法(MLR)及主成分回归分析法(PCA)对每个化合物的195个分子参数进行回归分析,分别建立了定量构效关系的最优预测模型;并用检验集检验了所建模型的预测能力。结果表明,多元线性回归法所建模型与主成分回归法所建模型相对比,发现逐步筛选法为最优建模方法。该方法所建模型统计结果良好(R=0.782,SEE=0.202),应用于检验集时结果也比较令人满意(R=0.764,SEP=0.114),模型表现出较强的可靠性和预测性。模型的建立和主要影响因素的确定有助于指导新型紫杉醇类似物药物的筛选和研发。     

新型靛红腙类合成大麻素质谱裂解规律研究

研究了新型靛红腙类合成大麻素在电子轰击(EI)和电喷雾(ESI)电离模式下的质谱裂解规律,并建立了可疑物中该类合成大麻素的鉴定方法。采用气相色谱-质谱联用(GC-MS)和液相色谱-高分辨质谱联用(LC-Q-Orbitrap/MS)技术,对5种新型靛红腙类合成大麻素(MDA-19 (BZO-HEXOXIZID),5C-MDA-19 (Pentyl MDA-19,BZO-POXIZID),CHM-MDA-19 (BZO-CHMOXIZID),4en-pentyl MDA-19(BZO-4en-POXIZID),5F-MDA-19 (5F-BZO-POXIZID))的主要碎片离子和碎裂过程进行分析,并对获得的质谱图进行解析,推测该类合成大麻素的EI-MS及ESI-MSⁿ碎裂规律。EI-MS可获得比ESI-MSⁿ更多的碎片离子用于该类合成大麻素的结构推断。碎片离子6,7和8对应的质荷比(m/z)118 (C₈H₈N⁺),132(C₈H₆NO<...     

硫代磷酸二乙酯类农药半抗原设计及抗体识别特性

通过分析硫代磷酸二乙酯类农药的结构特点, 设计并合成了系列半抗原; 采用活泼酯法将半抗原分别与牛血清蛋白(BSA)和卵清蛋白(OVA)偶联制备了系列免疫原和包被原; 通过免疫新西兰大白兔获得了相应抗硫代磷酸二乙酯类农药的类特异性抗体. 建立检测硫代磷酸二乙酯类农药的间接竞争酶联免疫分析(ELISA)方法, 分析探讨了免疫半抗原结构对抗体特性的影响, 并阐述了包被半抗原结构对ELISA灵敏度的影响规律. 结果表明, 手臂取代位置在苯环对位且手臂较短的免疫原具有较好的免疫效果, 同时异源包被可以显著提高ELISA方法的灵敏度. 由抗体PAb-H1和包被原H6-OVA建立的间接竞争ELISA方法可以同时检测7个广泛使用的有机磷农药, 其半抑制浓度(IC50)分别为蝇毒磷(0.013 mg/L)、对硫磷(0.348 mg/L)、喹硫磷(0.022 mg/L)、三唑磷(0.035 mg/L)、甲拌磷(0.751 mg/L)、除线磷(0.850 mg/L)及辛硫磷(1.301 mg/L), 最低检测限符合国内外相关有机磷药物最大允许残留限量标准(MRLS)的检测要求.     

Computer-assisted prediction of pesticide substructure using mass spectra

Mass spectral classifiers of 16 substructures that are present in basic structures of pesticides have been investigated to assist pesticide residues analysis as well as screening of pesticide lead compounds. Mass spectral data are first transformed into 396 features, and then Genetic Algorithm-Partial Least Squares (GA-PLS) as a feature selection method and Support Vector Machine (SVM) as a validation method are implemented together to get an optimization feature set for each substructure. At last, a statistical method which is AdaBoost algorithm combined with Classification and Regression Tree (AdaBoost-CART) is trained to predict the 16 substructures presence/absence using the optimization mass spectral feature set. It is demonstrated that the optimum feature sets can be used to predict the 16 pesticide substructures presence/absence with mostly 85-100% in recognition success rate instead of the original 396 features.     

机器学习方法用于二氢叶酸还原酶抑制剂的活性预测

分别采用支持向量学习机、人工神经网络、调节性逻辑回归和K-最临近等机器学习方法对761个二氢叶酸还原酶抑制剂建立了其活性分类预测模型. 采用组成描述符和拓扑描述符表征抑制剂的分子结构及物理化学性质, 使用Kennard-Stone方法进行训练集的设计, 并用Metropolis Monte Carlo模拟退火方法作变量选择. 结果表明, 支持向量学习机优于其它机器学习方法, 所得到的最优模型具有较好的预测结果, 其预测正确率为91.62%. 说明通过合适的训练集设计及变量选择, 支持向量学习机方法可以很好地用于二氢叶酸还原酶抑制剂的活性分类预测.     

A comparative study on feature selection methods for drug discovery

Feature selection is frequently used as a preprocessing step to machine learning. The removal of irrelevant and redundant information often improves the performance of learning algorithms. This paper is a comparative study of feature selection in drug discovery. The focus is on aggressive dimensionality reduction. Five methods were evaluated, including information gain, mutual information, a chi2-test, odds ratio, and GSS coefficient. Two well-known classification algorithms, Na?ve Bayesian and Support Vector Machine (SVM), were used to classify the chemical compounds. The results showed that Na?ve Bayesian benefited significantly from the feature selection, while SVM performed better when all features were used. In this experiment, information gain and chi2-test were most effective feature selection methods. Using information gain with a Na?ve Bayesian classifier, removal of up to 96% of the features yielded an improved classification accuracy measured by sensitivity. When information gain was used to select the features, SVM was much less sensitive to the reduction of feature space. The feature set size was reduced by 99%, while losing only a few percent in terms of sensitivity (from 58.7% to 52.5%) and specificity (from 98.4% to 97.2%). In contrast to information gain and chi2-test, mutual information had relatively poor performance due to its bias toward favoring rare features and its sensitivity to probability estimation errors.     

Determination of several pesticides in water by solid-phase extraction, liquid chromatography and electrospray tandem mass spectrometry

The analysis of pesticides in water samples is a problem of primary concern for quality control laboratories due to the toxicity level of these compounds and their public health risk. In order to evaluate the impact of pesticides in the Lisbon drinking water supply system, following the requirements of the European Union Directive 98/83/EC, we developed and validated an analytical method based on the combination of solid-phase extraction with liquid chromatography and tandem mass spectrometry. In this work, several pesticides were studied: imidacloprid, dimethoate, cymoxanil, carbendazime, phosmet, carbofuran, isoproturon, diuron, methidathion, linuron, pyrimethanil, methiocarbe, tebuconazole and chlorpyrifos. Several parameters of the electrospray source were optimized in order to get the best formation conditions of the precursor ion for each pesticide, namely capillary and extractor voltage, cone voltage, cone gas flow rate and desolvation gas flow rate. After optimization of the collision cell energy of the triple quadrupole, two different precursor ion-product ion transitions were selected for each pesticide, one for quantification and one for qualification, and these ions were monitored under time-scheduled multiple reaction monitoring (MRM) conditions. The selection of specific fragment ions for each pesticide guarantees a high degree of selectivity as well as additional sensitivity to quantify trace levels of these pesticides in water samples. This method showed excellent linearity ranges for all pesticides, with correlation coefficients greater than 0.9989. Determination limits (between 0.0041 and 0.0480 microg/L), precision (RSD <9.18%), accuracy and recovery studies in several water samples using solid-phase extraction were also performed.     

杭州老虎洞窑古陶瓷成分的化学计量学研究

用支持向量机算法研究了与杭州老虎洞古陶瓷有关的两个断源、断代问题。作为化学计量学的～种新型分类算法,支持向量机在小样本问题上表现出良好的泛化能力,与特征选择方法结合,可以有效处理样本少,特征多的问题。本研究综合利用支持向量机、特征选择算法和其它化学计量学算法研究了杭州凤凰山麓万松岭附近的古窑遗址和“传世哥窑”的断源、断代问题,证明老虎洞窑与郊坛下窑产品截然不同,万松岭附近地面收集瓷片样本是老虎洞窑宋代地层的瓷片滑落所致,而“传世哥窑”样品可能是老虎洞窑元代时的产品。实验表明：支持向量机算法与化学分析相结合可以成为研究古陶瓷断源和断代问题的一种新方法。