广义减秩—荧光法测定萘,菲,芘,芴四元混合物

提出了把主要成分分析用于校准和预报样本的新的减秩过程,并根据奇值分解所求的特征值所占其总和的百分比来确定主成分数。将该法应用于稠环芳烃萘、芴、菲、芘混合物荧光法的定量分析中,讨论了主成分数的影响,平均误差<8％,最大误差大约13％。     

有机磷农药构效关系的主成分分析-人工神经网络研究

采用主成分分析法对样本数据集进行预处理，将得到的新的样本数据集输入人工神经网络，对有机磷农药的毒性参数进行预报。研究结果表明，主成分分析－人工神经网络的预报精度优于单纯的人工神经网络。     

短波近红外光谱技术对葡萄酒中总糖含量快速测定的研究

采用短波近红外光谱技术结合偏最小二乘法(PLS),建立了葡萄酒中总糖含量的定量分析数学模型,讨论了光谱预处理方法和主成分数对PLS模型预报精度的影响.应用所建模型对预测集样本中总糖含量进行预报,结果令人满意.该方法方便快捷,并且具有较高的预报精度,可以用于葡萄酒中总糖含量的快速测定.     

校正变换矩阵法及其在多组分直接同时测定中的应用

将目标变换因子分析中自由浮动技术引入多元校准中，提出了一种新的多组分同时校准法———校正变换矩阵法。该法用于四组分氨基酸人工混合样品分析，结果令人满意，通过与传统目标变换因子分析法比较研究，表明该法用于组分光谱严重共线的波长范围时，其校准能力大大优于传统目标变换法。     

主成分-人工神经网络在近红外光谱定量分析中的应用

近红外光谱的主成分由非线性迭代偏最小二乘法（ＮＩＰＡＬＳ）求出。主成分作标准化处理后,作为Ｂ－Ｐ神经网络的输入结点进行非线性迭代。该法的优点是,充分利用了全光谱的数据,得到消除噪声后的最佳主成分,能建立非线性模型,Ｂ－Ｐ神经网络迭代时间显著缩短。用该法对大麦中的淀粉含量进行了定量分析研究。结果为：校准和预测的相关系数分别为０．９８１和０．９５３,校准和预测的相对标准偏差分别为１．７０％和２．４８％。     

主成分—人工神经网络在近红外光谱定量分析中的应用

近红外光谱的主成分由非线性迭代偏最小二乘法（ＮＩＰＡＬＳ）求出,主成分作标准化处理后,作为Ｂ－Ｐ神经网络的输入结点进行非线性迭代。该法的优点是,充分利用了全光谱的数据,得到消除噪声后的最佳主成分,能建立非线性模型,Ｂ－Ｐ神经网络迭代时间显著缩短,用该法对大麦中的淀粉含量进行了定量分析研究。结果为：校准和预测的相关系数分别为０．９８１和０．９５３,校准和预测的相对标准偏差分别为１．７０％和２．４８％     

玻璃标样结合硫内标归一定量技术在激光剥蚀-等离子体质谱分析硫化物矿物中的应用

以硬石膏矿物标样中Ca相对于S的灵敏度因子为基准,将玻璃标样中主量和痕量元素相对于Ca的灵敏度因子转换成元素相对于S的灵敏度因子,建立了多玻璃标样结合硫内标归一定量技术分析硫化物单矿物多元素的新方法。利用本方法分析了美国合成多金属硫化物矿物标样MASS-1中20种元素,主量元素分析结果的相对误差小于10%,痕量元素分析结果几乎都落在给定值±不确定度范围内。利用本方法对12个硫化物单矿物分析结果表明,绝大多数主量元素含量测定值的相对误差小于10%,且多数主量元素甚至优于以MASS-1为外标、内标归一定量法及内标校准法分析结果,而痕量元素与MASS-1校准结果较为一致。本方法克服了基体不匹配的问题,能比较准确地定量分析硫化物矿物中的主成分S,可用于定量校准硫化物矿物。     

主成分回归—催化动力学—掩蔽法同时测定钼和钨

利用掩蔽剂与催化同一指示反应的多种催化剂之间的掩蔽反应速率事程度的差异，提出通过一次动力学过程的测量，同时测定多种催化剂的方法。当掩蔽反应为快速反应或可逆反应时，该方法同样适用。以Ｈ２Ｏ２氧化ＫＩ为指示反应，柠檬酸作为掩蔽剂，借助微机直接记录碘离子选择电极在反应过程中的电位变化，以主成分回归作为建模和预报扔多元校正方法，对Ｍｏ和Ｗ进行了同时测定。     

主成分分析同时单点R滴定法研究

将主成分分析用于单点Ｒ滴定法中，同时测定了镍矿中Ｎｉ，Ｃｕ，Ｃｏ含量，讨论了方法原理，指定电位的选择，建立了主成分分析常数矩阵，对２０个模拟样和矿样进行了分析，均获得满意结果。     

基于料酒的三维同步荧光光谱模式识别研究

提出了一种应用同步荧光光谱技术无损快速鉴别料酒品牌的新方法.利用主成分分解法和小波变换法对料酒样品的同步荧光光谱信号进行了压缩处理,分别采用同步荧光光谱数据的第一主成分和小波细节系数为特征变量进行主成分分析和聚类分析,分类结果表明小波系数作为料酒的特征变量对料酒品牌分类正确率更高.利用偏最小二乘和径向基人工神经网络方法...     

A new strategy to improve the predictive ability of the local lazy regression and its application to the QSAR study of melanin‐concentrating hormone receptor 1 antagonists

In the quantitative structure‐activity relationship (QSAR) study, local lazy regression (LLR) can predict the activity of a query molecule by using the information of its local neighborhood without need to produce QSAR models a priori. When a prediction is required for a query compound, a set of local models including different number of nearest neighbors are identified. The leave‐one‐out cross‐validation (LOO‐CV) procedure is usually used to assess the prediction ability of each model, and the model giving the lowest LOO‐CV error or highest LOO‐CV correlation coefficient is chosen as the best model. However, it has been proved that the good statistical value from LOO cross‐validation appears to be the necessary, but not the sufficient condition for the model to have a high predictive power. In this work, a new strategy is proposed to improve the predictive ability of LLR models and to access the accuracy of a query prediction. The bandwidth of k neighbor value for LLR is optimized by considering the predictive ability of local models using an external validation set. This approach was applied to the QSAR study of a series of thienopyrimidinone antagonists of melanin‐concentrating hormone receptor 1. The obtained results from the new strategy shows evident improvement compared with the commonly used LOO‐CV LLR methods and the traditional global linear model. © 2009 Wiley Periodicals, Inc. J Comput Chem, 2010     

Combining the least correlation design, wavelet packet transform and correlation coefficient test to reduce the size of calibration set for NIR quantitative analysis in multi-component systems

The paper focuses on solving a common and important problem of NIR quantitative analysis in multi-component systems: how to significantly reduce the size of the calibration set while not impairing the predictive precision. To cope with the problem orthogonal discrete wavelet packet transform (WPT), the least correlation design and correlation coefficient test (r-test) have been combined together. As three examples, a two-component carbon tetrachloride system with 21 calibration samples, a two-component aqueous system with 21 calibration samples, and a two-component aqueous system with 41 calibration samples have been treated with the proposed strategy, respectively. In comparison with some previous methods based on much more calibration samples, the results out of the strategy showed that the predictive ability was not obviously decreased for the first system while being clearly strengthened for the second one, and the predictive precision out of the third one was even satisfactory enough for most cases of quantitative analysis. In addition, all important factors and parameters related to our strategy are discussed in detail.     

A simple and readily integratable approach to toxicity prediction

A simple, highly extensible computational strategy to assess compound toxicity has been developed with the premise that a compound's toxicity can be gauged from the toxicities of structurally similar compounds. Using a reference set of 13645 compounds with reported acute toxicity endpoint dose data (oral, rat-LD(50) data normalized in mg/kg), a generic utility which assigns a compound the average toxicity of structurally similar compounds is shown to correlate well with reported values. In a leave-one-out simulation using the requirement that at least one structurally similar member in a "voting consortium" is present within a reference set, the strategy demonstrates a predictive correlation (q wedge 2) of 0.82 with 57.3% of the compounds being predicted. Similar leave-one-out simulations on a set of 1781 drugs demonstrate a q wedge 2 of 0.74 with 51.8% of the compounds being predicted. Simulations to optimize similarity cutoff definitions, consortium member size, and reference set size illustrate that a significant improvement in the quality and quantity of predictions can be obtained by increasing the reference set size. Similar application of the strategy to subchronic and chronic toxicity data should be possible given reasonably sized reference sets.     

Prediction of protein structural classes by a new measure of information discrepancy

Since it was observed that the structural class of a protein is related to its amino acid composition, various methods based on amino acid composition have been proposed to predict protein structural classes. Though those methods are effective to some degree, their predictive quality is confined because amino acid composition cannot sufficiently include the information of protein sequences. In this paper, a measure of information discrepancy is applied to the prediction of protein structural classes; different from the previous methods, this new approach is based on the comparisons of subsequence distributions; therefore, the effect of residue order on protein structure is taken into account. The predictive results of the new approach on the same data set are better than those of the previous methods. As to a data set of 1401 sequences with no more than 30% redundancy, the overall correctness rates of resubstitution test and Jackknife test are 99.4 and 75.02%, respectively, and to other data sets the similar results are also obtained. All tests demonstrate that the residue order along protein sequences plays an important role on recognition of protein structural classes, especially for alpha/beta proteins and alpha+beta proteins. In addition, the tests also show that the new method is simple and efficient.     

Computational studies on N-phenyl pyrrole derivatives as MmpL3 inhibitors in Mycobacterium tuberculosis

The fight against tuberculosis (TB) is a time immemorial one and the emergence of new drug resistant strains of Mycobacterium tuberculosis keeps throwing new challenges to the scientific community immersed in finding mechanisms to control this dreaded disease. Computer aided drug designing (CADD) is one of the several approaches that can assist in identifying the potent actives against Mycobacterium. In this work, a series of 109 known Mycobacterial membrane proteins large 3 (MmpL3) inhibitors were pooled and atom based 3D QSAR analysis was performed to understand the structural features essential for inhibitory activity against the MmpL3, known to be a key player in transporting substances critical for cell wall integrity of Mycobacterium. The data set employed was randomly split into training set and test set molecules. The training set of 74 molecules was used to derive CoMFA and CoMSIA models that were statistically reliable (CoMFA: q²_loo = 0.53; r²_ncv = 0.93 and CoMSIA: q²_loo = 0.60; r²_ncv = 0.93). The derived models also exhibited good external predictive ability (CoMFA: r²_pred = 0.78 and CoMSIA: r²_pred = 0.79). The results are quite encouraging and information derived from these analyses was applied to design new molecules. The designed molecule showed appreciable predicted activity values and reasonably good ADMET profile. The strategy used in designing new molecules can be pursued in the hunt for new chemical entities targeting MmpL3, expanding the existing arsenal against TB.     

机器学习方法用于选择性环氧化酶-2抑制剂活性预测模型的建立

与传统的非甾体类消炎药相比,选择性环氧化酶-2抑制剂具有无胃肠道粘膜损伤,溃疡和肾功能障碍等严重的副作用,设计选择性环氧化酶-2抑制剂具有重要意义。本文用支持矢量学习机和神经网络两种机器学习方法建立选择性环氧化酶-2抑制剂的活性预测模型,以期为选择性环氧化酶-2抑制剂药物的合成提供先导化合物。我们将467个环氧化酶-2抑制剂用Kennard-Stone方法分为训练集,验证集和独立测试集,对每一抑制剂分子我们计算了463个包含组成描述符和拓扑描述符的分子描述符来表征其分子结构,并通过F-Score方法选取最重要的分子描述符用于分类模型的建立。结果表明,SVM方法通过变量筛选后具有很好的预测能力,其预测正确率达到93.30%。     

The evaluation of probabilistic classification methods : Part 1. A monte carlo study with ALLOC

Probabilistic classification (i.e., classification of individuals into one of several groups by assigning probabilities of classification to each individual) is desirable when the main interest is in individuals rather than the whole group. The evaluation of probabilistic assignments is described in detail, including statistical features such as measures for the sharpness of the classification, the predictive ability and the reliability of the probability values. In a simulation study, the influence of the objects—variable ratio and the interclass distance on the results was examined for the training data themselves (resubstitution method), an independent test set, and a pseudo-independent test set created from the training set (leave-one-out method). The results indicate that the leave-one-method can often be used instead of an independent test set. In many cases, the assignments cited as probabilities are not probabilities at all, because the classification system is too over-confident.     

On further application of r as a metric for validation of QSAR models

Validation is a crucial aspect for quantitative structure–activity relationship (QSAR) model development. External validation is considered, in general, as the most conclusive proof of predictive capacity of a QSAR model. In the absence of truly external data set, external validation is usually performed on test set compounds, which are members of the original data set but not used in model development exercise. In the case of small data sets, QSAR researchers experience problem in model development due to the fact that the developed models may be less reliable on account of the small number of training set compounds and such models may also show poor external predictability because the models may not have captured all necessary features required for the particular structure–activity relationships. The present paper attempts to show that ‘true r_(LOO)’ statistic calculated based on the model derived from the undivided data set with application of variable selection strategy at each cycle of leave‐one‐out (LOO) validation may reflect external validation characteristics of the developed model thus obviating the requirement of splitting of the data set into training and test sets. This approach may be helpful in the case of small data sets as it uses all available data for model development and validation thus making the resulting model more reliable. Copyright © 2009 John Wiley & Sons, Ltd.