Quantitative structure-activity variable selection

Quantitative Structure-Activity Relationship models are used increasingly in chemical data mining and combinatorial library design [5, 6]. For example, three-dimensional (3-D) stereoelectronic pharmacophore based on QSAR modeling was used recently to search the National Cancer Institute Repository of Small Molecules [7] to find new leads for inhibiting HIV type 1 reverse transcriptase at the nonnucleoside binding site [8]. A descriptor pharmacophore concept was introduced by us recently [9] on the basis of variable selection QSAR the descriptor pharmacophore is defined as a subset of... 

Rogers, D. Hopfingee, A.J. Application of genetic function approximation to quantitative structure-activity relationships and quantitative structure-property relationships. J. Chem. Inf. Comput. Sci. 1994, 34, 854-866. Kubinyi, H. Variable selection in QSAR studies. 1. An evolutionary algorithm. Quantum Struct.-Act. Relat. 1994, 13, 285-294. 

Kubinyi, H., Variable selection in QSAR studies, Quantitative Structure-Activity Relationships 13, 285-294, 1994. 

When compounds are selected according to SMD, this necessitates the adequate description of their structures by means of quantitative variables, "structure descriptors". This description can then be used after the compound selection, synthesis, and biological testing to formulate quantitative models between structural variation and activity variation, so called Quantitative Structure Activity Relationships (QSARs). For extensive reviews, see references 3 and 4. With multiple structure descriptors and multiple biological activity variables (responses), these models are necessarily multivariate (M-QSAR) in their nature, making the Partial Least Squares Projections to Latent Structures (PLS) approach suitable for the data analysis. PLS is a statistical method, which relates a multivariate descriptor data set (X) to a multivariate response data set Y. PLS is well described elsewhere and will not be described any further here [42, 43]. 

Marengo, E., Carpignano, R., Savarino, P. and Viscardi, G. (1992). Comparative Study of Different Structural Descriptors and Variable Selection Approaches Using Partial Least Squares in Quantitative Structure-Activity Relationships. Chemom.Intell.Lab.Syst., 14,225-233. 

Waller, C.L. and Bradley, M.P. (1999). Development and Validation of a Novel Variable Selection Technique with Application to Multidimensional Quantitative Structure-Activity Relationship Studies. J.Chem.lnfComput.ScL, 39,345-355. 

In order to optimise the in vitro profile, we focused our attention on the nature of the substituent at N-1 and a quantitative structure-activity study was performed on a series of N-1 alkyl derivatives. After selection of variables, the affinity for the CCK-B receptor was related to the calculated values of both lipophilicity [26] and molar refractivity [27] of the substituent and the following equation was derived using PLS analysis implemented in program GOLPE [28] (all parameters are referred to the substituents at N-1) ... 

Deconinck et al. (53) used CART in a quantitative structure-activity relationship context on an intestinal absorption data set of 141 drug-hke molecules. Many theoretical molecular descriptors were calculated and used as explanatory variables (X matrix). The considered response (y) was the percentage human intestinal absorption of the compounds. The total sum of squares of the response values about the mean of the node was applied as impurity measure. From all descriptors, only two were chosen to describe and predict the intestinal absorption, and this resulted in three terminal nodes. However, the tree thus obtained did not allow dehning classes with a limited absorption range, and therefore more complex trees were evaluated. Finally, a tree with 11 terminal nodes was selected. The absorption of the molecules was divided into five (absorption) classes. Each terminal node was labeled with one or two class symbols. From an external test set, three out of 27 molecules were wrongly classified (11.1%). 

How can parsimonious models be constructed There are several possible approaches, however in this chapter a combination of data compression and variable selection will be used. Data compression achieves parsimony through the reduction of the redundancy in the data representation. However, compression without involving information about the dependent variables will not be optimal. It is therefore suggested that variable selection should be performed on the compressed variables and not on the original variables which is the usual strategy. Variable selection has been applied with success in fields such as analytical chemistry [1-4], quantitative structure-activity relationships (QSAR) [5-8] and analytical biotechnology [9-11]. 

Waller, C. L. and Bradley, M. P. (1999) Development and validation of a novel variable selection technique with application to multidimensional quantitative structure-activity relationship studies. J. Chem. Inf. Comput. Sci. 39, 345-355. 

Initially, 16 molecular descriptors were selected as possible independent variables for quantitative structure-activity relationships (OSAR). They were ... 

In summary, the support vector machine (SVM) and partial least square (PLS) methods were used to develop quantitative structure activity relationship (QSAR) models to predict the inhibitory activity of nonpeptide HIV-1 protease inhibitors. Cenetic algorithm (CA) was employed to select variables that lead to the best-fitted models. A comparison between the obtained results using SVM with those of PLS revealed that the SVM model is much better than that of PLS. The root mean square errors of the training set and the test set for SVM model were calculated to be 0.2027, 0.2751, and the coefficients of determination (R2) are 0.9800, 0.9355 respectively. Furthermore, the obtained statistical parameter of leave-one-out cross-validation test (Q ) on SVM model was 0.9672, which proves the reliability of this model. Omar Deeb is thankful for Al-Quds University for financial support. 

D QSAR analyses based on similarity matrices offer a valuable new tool for the quantitative description of structure-activity relationships. Also hydrophobic fields and interaction fields with different probe atoms may be implemented, like in CoMFA studies. It is hoped that the preliminary results [1064, 1065] stimulate active research in this field to achieve further methodological improvements. Several CoMFA-inherent problems apparently do not arise in the molecular similarity matrices approach, e.g. the cut-off selection, a proper grid spacing, and the elimination of variables having low variance. 

Details of the TDI and MDI foam model systems have been previously published [2]. The models require the use of mono-functional reactants that are quantitatively analysed to correlate structure-activity relationships for various classes of catalysts. A realistic thermal profile is produced through the imposition of an external exotherm. Urethane, urea, allophanate and biuret reaction products are quantified by liquid chromatographic analysis of quenched reaction samples. The models effectively account for such nonideal conditions as reactant depletion at variable rates, temperature and concentration-dependent catalyst activity, and catalyst selectivity as a function of isocyanate distribution. 

---