Boosting support vector regression in QSAR studies of bioactivities of chemical compounds.

In this paper, boosting has been coupled with SVR to develop a new method, boosting support vector regression (BSVR). BSVR is implemented by firstly constructing a series of SVR models on the various weighted versions of the original training set and then combining the predictions from the constructed SVR models to obtain integrative results by weighted median. The proposed BSVR algorithm has been used to predict toxicities of nitrobenzenes and inhibitory potency of 1-phenyl[2H]-tetrahydro-triazine-3-one analogues as inhibitors of 5-lipoxygenase. As comparisons to this method, the multiple linear regression (MLR) and conventional support vector regression (SVR) have also been investigated. Experimental results have shown that the introduction of boosting drastically enhances the generalization performance of individual SVR model and BSVR is a well-performing technique in QSAR studies superior to multiple linear regression.     

Exploring QSARs for antiviral activity of 4-alkylamino-6-(2-hydroxyethyl)-2-methylthiopyrimidines by support vector machine

The support vector machine, which is a novel algorithm from the machine learning community, was used to develop quantitative structure activity relationship models to predict the antiviral activity of 4-alkylamino-6-(2-hydroxyethyl)-2-methylthiopyrimidines. The genetic algorithm was employed to select the variables that resulted in the best-fitted models. A comparison between the obtained results using support vector machine with those of multiple linear regression revealed that support vector machine model was much better than multiple linear regression. The root mean square errors of the training set and the test set for support vector machine model were calculated to be 0.102 and 0.205, and the correlation coefficients (r2) were 0.956 and 0.852, respectively. Furthermore, the obtained statistical parameter of leave-one-out (LOO) and leave-group-out (LGO) cross-validation test on support vector machine model were 0.893 and 0.881, respectively, which prove the reliability of this model. The results suggest that branching, volume and lipophilicity are the main independent factors contributing to the antiviral activities of the studied compounds.     

Insights into the structural requirement of 6-nitroquinolone-3-carboxylic acids as antimycobacterial agent: chemometric approaches

Quantitative relationship between structures of 6-nitroquinolone-3-carboxylic acids and their antimycobacterial activity was investigated by different chemometric methods namely multiple linear regression, sequential multiple linear regression, partial least square, and machine learning support vector machine. The obtained models were able to describe about 81–90% of the variance in the experimental activity of molecules in the training set. The accuracy and predictability of the proposed models were illustrated using various evaluation techniques like internal and external validation. The best quantitative structure–activity relationship model reveals that van der Waals volume and Sanderson electronegativity as the most influencing atomic properties in the structures of the carboxylic acid derivatives. The proposed model may provide a better understanding of the antimycobacterial activity of 6-nitroquinolone-3-carboxylic acid analogs and can be used as guidance for proposition of new anti-tubercular agents.     

Regression methods for developing QSAR and QSPR models to predict compounds of specific pharmacodynamic, pharmacokinetic and toxicological properties

Quantitative structure-activity relationship (QSAR) and quantitative structure-property relationship (QSPR) models have been extensively used for predicting compounds of specific pharmacodynamic, pharmacokinetic, or toxicological property from structure-derived physicochemical and structural features. These models can be developed by using various regression methods including conventional approaches (multiple linear regression and partial least squares) and more recently explored genetic (genetic function approximation) and machine learning (k-nearest neighbour, neural networks, and support vector regression) approaches. This article describes the algorithms of these methods, evaluates their advantages and disadvantages, and discusses the application potential of the recently explored methods. Freely available online and commercial software for these regression methods and the areas of their applications are also presented.     

土壤水分特征曲线模型参数识别的多邻域粒子群算法

Van Genuchten模型(简称VG模型)是目前运用最为广泛的土壤水分特征曲线模型，提出适宜的优化算法进行模型参数识别也是一个非常重要的研究方向。针对标准的粒子群算法易陷入局部最优的缺点，给出了一种多邻域粒子群算法，可以有效地克服粒子群算法易陷入局部最优的缺点，并利用该算法对VG模型参数进行识别，最后用所求解的参数对不同类型土壤持水性能进行了试验。数值实验结果表明，多邻域粒子群算法能够有效地应用于VG模型的参数识别，与其它算法相比在性能和精度上都有所提高，而且对参数的取值范围也可以较大地放宽。因此，多邻域粒子群算法可以作为VG模型参数识别的一种新方法。     

Chemometrics‐Based Approach to Feature Selection of Chromatographic Profiles and its Application to Search Active Fraction of Herbal Medicine

In our previous report (J Pharmaceut Biomed 56 (2011) 443–447), a support vector machine (SVM)‐based pharmacodynamic model was established for predicting active fractions of herbal medicines (HMs), where information contents embedded in the chromatograms of the fractions were represented with the peak areas. However, in this representation the global characteristics of the chromatograms were completely missed, which is definitely contrary to the global and holistic views in theories of HMs and undoubtedly reduce the success rate of this model. To deal with the challenge, two chemometrics methods, that is, minimum redundancy maximum relevance (mRMR) and particle swarm optimizer (PSO), were applied in this article for feature selection of the whole chromatograms, and the PSO was also used to tune the SVM parameters. As a case, a sample HM, that is, Xiangdan injection, was investigated. The predictive accuracy was fully evaluated and compared with those by other popular and reported methods. Furthermore, the confirmation on the independent predicting set exhibited that the predicted bioactivities were well consistent with the experimental values. The important potential application of the present model is to be extended to help search active fractions of other HMs.     

QSAR models for isoindolinone-based p53-MDM2 interaction inhibitors using linear and non-linear statistical methods

The design and optimization of p53-MDM2 interaction inhibitors has attracted a great deal of interest in the development of new anticancer agents. Systematical 2D-QSAR studies on 98 isoindolinone-based p53-MDM2 interaction inhibitors were carried out using linear and the non-linear mathematical methods. At first, a forward stepwise-multiple linear regression model (FS-MLR) was proposed with reasonable statistical parameters (R(2)(train) =0.881, Q(2)(loo) =0.847, R(2)(test) =0.854). Then, enhanced replacement method-multiple linear regression (ERM-MLR) and support vector machine regression (SVMR) were applied to set up more accurate models (ERM-MLR: R(2)(train) =0.914, Q(2)(loo) =0.894 and R(2)(test) =0.903; SVMR: R(2)(train) =0.924, Q(2)(loo) =0.920 and R(test) (2) of 0.874). Furthermore, the reliability and application value of the ERM and SVMR model was also validated in virtual screening through receiver operating characteristic studies.