钙调蛋白抑制剂预测模型研究

选用30个结构多样的caM抑制剂分子作为数据集,采用多元线性回归(MLR)方法及主成分回归分析(PCA)方法对每个化合物的194个分子参数进行回归分析,分别建立了各自的最优预测模型.结果表明:多元线性回归分析方法所建模型与主成分回归所建模型相对比,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(R2=0.952,SEE为0.289),应用于检验集时结果也比较令人满意(R2=0.941,SEP为0.295),模型表现出较强的可靠性和预测性.     

P2Y12受体拮抗剂定量构效关系的研究

P2Y12受体拮抗剂是一类重要的抗血小板药物,研究分子活性与其结构参数的关系,对于合成新的P2Y12受体拮抗剂具有一定指导作用.选用178个结构多样的P2Y12受体拮抗剂分子作为数据集,随机选取了143个P2Y12受体拮抗剂作为训练集,剩余分子作为检验集.采用多元线性回归(MLR)方法和主成分回归分析(PCA)方法对每个分子的636个分子参数进行线性回归分析.MLR所建模型的结果为:训练集R2=0.800,检验集R2=0.834;PCA模型结果为:训练集R2=0.545,检验集R2=0.665.相比之下MLR法所建模型具有良好的预测性和可靠性.通过模型分析,确定了影响分子活性的关键因素.以上模型对筛选和合成新型高效P2Y12受体拮抗剂提供了一定理论指导.     

吡唑啉酮类衍生物对c-Met抑制活性的特征结构研究

肝细胞生长因子(HGF)/酪氨酸蛋白激酶(c-Met)介导的细胞信号是导致肿瘤细胞产生和转移的主要途径之一。c-Met抑制剂能够阻断HGF/c-Met信号,抑制人类肿瘤的转移发生。本文选用77个结构多样的吡唑啉酮类衍生物作为c-Met抑制剂分子的数据集,随机选取其中16个分子作为检验集,其余作为训练集,采用多元线性回归(MLR)和主成分回归分析(PCA)法对每个分子的648个参数进行回归分析,分别建立定量构效关系的最优预测模型。结果表明,多元线性回归中的逐步筛选法是最佳的建模方法,其所建模型统计结果良好(R2=0.81,SEE=0.37),应用于检验集的结果也较理想(R2=0.83,SEP=0.42),模型可靠性和预测能力较强,能直观反映影响活性的主要因素。此模型的确立有助于指导新型高效c-Met抑制剂药物的筛选和开发。     

HCV复制抑制剂的定量构效关系的研究

选用60个结构多样的HCV复制抑制剂分子作为数据集,随机选择其中46个分子作为训练集,剩余14个分子作为验证集.采用多元线性回归(MLR)和主成分分析(PCA)方法对每个分子的646个理化和结构参数进行了线性回归分析,并分别建立各自的最优模型.结果表明MLR中的逐步和向前法所建模型最佳,模型结果为:训练集R2=0.827,验证集R2=0.850,模型能够直观地反映影响化合物活性的主要因素.该模型将有助于筛选和开发新的HCV复制抑制.     

噻吩和三唑酮类Chk1抑制剂的特征结构研究

检验点激酶1（Chk1）是一种丝氨酸/苏氨酸蛋白激酶,可使受损癌细胞发生细胞周期阻滞并促进DNA损伤修复,降低放射治疗对肿瘤细胞的杀伤作用。抑制Chk1活性可消除细胞周期阻滞,阻碍肿瘤细胞的自我修复,导致其有丝分裂紊乱最终凋亡。本文采用多元线性回归(MLR)和主成分回归分析(PCA)法,对164个噻吩和三唑酮类Chk1抑制剂分子的698个参数进行回归分析,分别建立定量构效关系预测模型。结果表明,多元线性回归中的forward法（向前选择法）是最佳建模方法,所建模型的预测能力较强（训练集R=0.851,检验集     

基于分子参数的药物小肠吸收预测模型

选择100个化合物作为数据集,随机选取其中80个为训练集,其他分子为验证集,并为每个化合物分子计算了30个参数.通过采用五种不同多元线性回归分析方法对其训练模拟,建立了数学模型,并用验证集检验了所建模型的预测能力.结果发现向后筛选法为最优小肠吸收建模方法.由该法所建模型的统计结果良好(R2>0.80),应用于验证集时也表现出较强预测能力.该模型确定了对小肠吸收影响较大的分子参数,有助于指导进一步的新药筛选和开发.     

噻吩和三唑酮类Chk1抑制剂的特征结构研究

检验点激酶1(Chk1)是一种丝氨酸/苏氨酸蛋白激酶,可使受损癌细胞发生细胞周期阻滞并促进DNA损伤修复,降低放疗对肿瘤细胞的杀伤作用。抑制Chk1活性可消除细胞周期阻滞,阻碍肿瘤细胞的自我修复,导致其有丝分裂紊乱最终凋亡。本文采用多元线性回归(MLR)和主成分回归分析(PCA)方法,对164个噻吩和三唑酮类Chk1抑制剂分子的698个参数进行回归分析,分别建立定量构效关系预测模型。结果表明,多元线性回归中的逐步回归(stepwise)是最佳建模方法,所建模型的预测能力较强(训练集R=0.85,检验集R=0.88),运用拓扑电荷指数、原子核片段等20个参数直观反映了影响活性的主要因素。此模型的确立对设计和开发新型高效Chk1抑制剂具有指导意义。     

含氟取代苯类化合物定量结构与毒性关系研究

本文采用了按氢分类的分子电距矢量(H-MEDV)对19个含氟取代苯类化合物进行了结构表征。采用多元线性回归(MLR)方法建立了含氟取代苯类化合物的定量结构-毒性关系(QSAR)的5变量模型;上述模型对19个含氟取代苯类化合物毒性预测与实验值较好的吻合,其复相关系数(R)为0.918,标准偏差(SD)为0.382,结果表明所建模型具有较好的稳定性和预测能力。     

苯环己胺类药物的QSAR研究

采用量子化学密度泛函DFT/B3LYP方法,计算了30个苯环己胺(PCA)类化合物的电子结构参数,结合Chemoffice8.0软件计算的基于距离的拓扑参数,用逐步多元线性回归法建立了其结构参数与药物活性之间的定量结构-活性关系(QSAR).所建模型的复相关系数R=0.9250;用留一法(leave-one-out,L...     

分子电性距离矢量用于三嗪类化合物的构效关系研究

采用分子电性距离矢量(Molecular Electronegativity Distance Vector,MEDV)表征了三嗪类化合物的分子结构,并运用多元线性回归(Multiple Linear Regression,MLR)建立了该类化合物结构与其发光菌和大型蚤毒性的定量结构-毒性相关(Quanti-tative Structure-Toxicity Relationship,QSTR)模型,同时采用留一法交互检验对所建模型进行了分析和验证,建模计算值的相关系数R分别为0.970和0.952,留一法交互检验预测值的相关系数RLOO分别为0.917和0.921,并进一步阐述了结构与毒性之间的关系。结果表明,三嗪环上π电子离域程度减小有利于毒性增加,侧链N上取代基数目增加,化合物毒性减小。为进一步预测该类化合物的毒性,进行药物筛选提供了有效的理论依据。     

也谈计算机在测试数据统计处理中的应用——与秦建侯等同志商榷

本文对多元回归分析及其计算机程序进行了进一步研究.指出了《分析试验室》1989年第3期所载回归程序疏漏之处,并指出该程序采用回归系数简化计算法具有局限性。为此提出了另一种逐步回归程序,对一些经典数据进行回归分析,得到满意结果。     

A new computer program for QSAR-analysis: ARTE-QSAR

A new computer program has been designed to build and analyze quantitative-structure activity relationship (QSAR) models through regression analysis. The user is provided with a range of regression and validation techniques. The emphasis of the program lies mainly in the validation of QSAR models in chemical applications. ARTE-QSAR produces an easy interpretable output from which the user can conclude if the obtained model is suitable for prediction and analysis.     

Principal component multinomial regression and spectrometry to predict soil texture

The mineral particles are classified in different textural classes according to their size. Reflectance spectrometry and spectra can be valid instruments to classify the soils according to their texture. This is possible using different statistical methods, for example, discriminant analysis. However, other multivariate methods, like multinomial logistic regression, can be used, but the presence of multicollinearity among explicative variables could affect the estimation of the parameters. The solution proposed to remedy this problem is an alternative way to apply the multinomial logit model. To evaluate its performances, we compare the results with both classical multinomial logit and discriminant analysis ones. Copyright © 2015 John Wiley & Sons, Ltd.     

Enthalpy Relaxation in Glasses: Regression analysis of integral DSC data

A new method of calculation of parameters of enthalpy relaxation models is proposed. Regression analysis treatment compares the experimental and calculated values of relaxation enthalpy. The experimental values of relaxation enthalpy are obtained by numerical integration of the difference between the two DSC curves. Contrary to the overall shape of the DSC curve the integral values are not affected by particular heat flow conditions during the DSC experiment. The Narayanaswamy's numerical model based on the Kohlrausch—William—Watts relaxation function was used to calculate the theoretical values of relaxation enthalpy. The application of the proposed method on the DSC experimental data of enthalpy relaxation of As₂Se₃ is shown.This revised version was published online in November 2005 with corrections to the Cover Date.     

Temperature dependence of adsorption parameters of n-butanol and n-valeric acid at their adsorption on mercury electrode from aqueous solutions

Adsorption parameters that enter into the Frumkin isotherm and the model of two parallel capacitors are calculated based on the data of the regression analysis of differential capacitance curves of a mercury electrode in solutions of n-butanol and n-valeric acid at 25, 50, and 75°C. The analysis of the temperature dependences of these parameters allowed free energies, entropies, and enthalpies of adsorption to be found. It is shown that the hydrophobic effect, which is associated with the increase in the enthalpy of liquid water when adsorbate molecules leave it, makes a substantial contribution into the adsorption free energy of studied compounds.     

Data fusion in multivariate calibration transfer

We report the use of stacked partial least-squares regression and stacked dual-domain regression analysis with four commonly used techniques for calibration transfer to improve predictive performance from transferred multivariate calibration models. The predictive performance from three conventional calibration transfer methods, piecewise direct standardization (PDS), orthogonal signal correction (OSC) and model updating (MUP), requiring standards measured on both instruments, was significantly improved from data fusion either by stacking of wavelet scales or by stacking of spectral intervals, as demonstrated by transfer of calibrations developed on near-infrared spectra of synthetic gasoline. Stacking did not produce as significant an improvement for calibration transfer using a finite impulse response (FIR) filter, but application of SPLS regression to FIR-transferred spectra improves predictive performance of the transferred model.     

High-temperature superconducting thin films: A multivariate statistical approach

Recent investigations have indicated that high-temperature superconducting thin films produced by chemical vapour deposition possess interesting properties. The present study reveals a new aspect of this trend, i.e. detection of similarities or dissimilarities of these materials with respect to their technological and superconducting properties, using some multivariate statistical approaches like cluster, principle components and multiple regression analysis. This approach adds to the conventional knowledge of fabrication parameters and superconducting properties by establishing the number and type of conditional factors responsible for these properties or technologies, correlation of variables (properties) within a certain conditional factor, similarities between film-producing teams even with the use of different technological parameters etc. Multivariate statistics make it possible to detect some quantitative links between superconducting and fabrication parameters and to suggest certain patterns of high-temperature superconducting thin films possessing common features.     

Measurement of pork freshness using potentiometric sensor

This study evaluated pork freshness using potentiometric solid-state electrodes in order to detect chemical indices such as reduced compounds, organic compounds and sulfides, which are produced during the initial stage of putrefaction in meat. Pt, CuS and Ag₂S electrodes selected as solid-state electrodes have, respectively, been used to detect the organic compounds (regarded as chemical indices of deterioration in meat freshness). The outputs of these electrodes have been analyzed by principal component analysis (PCA) and multiple regression analysis (MRA) in order to find the correlation with the results of viable bacterial counts. By using the potentiometric sensor, the pork freshness was evaluated and the PCA and MRA corresponded to the degree of bacterial increases more simply and rapidly than other methods such as viable bacterial counts or a biosensor.