Four-descriptors model

For selection of a model based on SC, some of the arguments mentioned above recommend the four-descriptor model, such as its minimal differences Sqy - S, and its maximal F. Figure 7.10 is a scatterplot of the calculated vs. experimental boiling points for this model, again containing values from both resubstitution and... 

The data containing 324 descriptor values of 88 molecules was given as an input to VSMP program, to build models based on three and four descriptors, keeping the interdescriptor correlation below 0.75. The best three-descriptors model, Eq. 80, was based on descriptors 254 (atomic type E-state index), 311 (AlogP98), and 320 (2D Van der Waals surface area) with a correlation coefficient, r, of 0.8425, and the cross-validated correlation coefficient, q, of 0.8239. The correlation coefficients of the other two VSMP models, Eqs. 81 and 82 were 0.8411 and 0.8329, respectively. Significantly, the descriptors 254 and 311 were selected in all the best three-descriptors models of VSMP. The three descriptors, in the models 80, 81, and 82 were 320, 144 (Kappa shape index of order 1), and 30 (topological Xu index), respectively. 

Xu J, Chen B, Zhang Q et al. (2004) Prediction of refractive indices of linear polymers by a four descriptor QSPR model. Polymer 45 8651-8659... 

Recently, Wichmann et al. [47] applied several COSMO-RS cr-moments as descriptors to model BBB permeability. The performance of the log BB model was reasonable given only four descriptors were applied n — 103, r2 = 0.71, RMSE = 0.4, LOO q2 — 0.68, RMSEtest = 0.42. The COSMO-RS cr-moments were obtained from quantum chemical calculations using the continuum solvation model COSMO and a subsequent statistical decomposition of the resulting polarization charge densities. 

Although there is a strong negative correlation between partition coefficient and aqueous solubility (Hansch et al., 1968 Chiou et al., 1977), and a strong positive correlation between % and molecular volume (Dearden et al., 1988), the use of the partial least squares (PLS) method in this study allows the simultaneous use of intercorrelated descriptors. Nevertheless, the use of four descriptors to model the bioconcentration factor of only 11 compounds contravenes the Topliss and Costello (1972) rule, and renders the QSAR of dubious validity. 

Statistical analysis showed that, although the separate QSAR models and PCM gave similar predictive ability, interpretation of the PCM model was more reliable. After variable selection, each of the four QSAR models retained different sets of descriptors, which made it difficult to find determinants for the receptor-subtype selectivity of the compounds. In contrast, interpretation of the PCM model was straightforward, and the selectivity of the compounds could be explained on the basis of a relatively small number of cross terms [15]. 

Elguero et al. [141] have reduced Palm s analogous tetraparametric model for the multiple correlation of solvent effects [cf. Eq. (7A8) in Section 7.7] to a triparametric one, with two factors explaining 94% of the data variance given in an original set of four descriptors [Y, P, E, and B of Eq. (7-48) in Section 7.7] for 51 solvents. 

Roy and Leonard [183] have also presented QSAR models for the HIV-1 RT inhibitory activity of the thiazolidinones listed in Tables 16 and 17 along with some more similar analogues (Table 18) [184-187] using the hydrophobic-ity and molar refractivity, quantum chemical and topological and indicator parameters as descriptors. In this, the 3-pyridyls/phenyls (Tables 16 and 17) and 3-(pyrimidin-2-yls) (Table 18) [186] have become part of the dataset. Additionally, four compounds with the thiazolidin-4-thione nucleus have been included in the dataset. In Fujita-Ban [188] and mixed (Hansch and Fujita-Ban) approaches 7 to 17 descriptor models have been discovered for the cytopathicity effect (EC50) and cytotoxic effect (CC50) of the compounds. The following equations show the minimum descriptor models for each activity from this study. 

Later, Katritzky et al. [88] reported the CODESSA PRO-based QSAR approach to understand the importance of the charge distribution and the H-bonding interaction of both symmetrical and imsymmetrical tetrahy-dropyrimidinones (22). Four QSAR models (23-26) were developed [88] for the enzyme inhibition data studied by Delucca et al. [89]. The first QSAR contained a topological descriptor based on Shannons expression of information content ( SIC) [90] as a term representing molecular size and shape. 

This way, the redundancies or double counting of models are avoided even with the price of jumping some intermediate models, as are the four-descriptors endpoints. The results of Table 3.50 are interpreted in the sense of establishing the minimum of three path hierarchies, and then compared at the global level note that more than three minimum paths will produce redundant information. Accordingly, the minimum paths, for LoSMoC and BraS cases (i)/(ii) separately, are (Putz Duda , 2013b) ... 

Models developed with selected subsets of descriptors, instead of all possible descriptors, can be more accurate and robust. In order to select adequate descriptors for each of the four classes of protons, genetic algorithms (GA) were used, and the results were compared with those obtained when all the descriptors were used. 

---