Comparative study of different structural descriptors and variable selection approaches using partial least squares in quantitative structure-activity relationships

Quantitative structure-activity relationship (QSAR) studies have been performed on piperidine derivatives (n = 119) as CCR5 antagonists. The whole data set was divided into a training set (75% of the dataset) and a test set (remaining 25%) on the basis of K-means clustering technique. Models developed from the training set were used to assess the predictive potential of the models using test set compounds. Initially classical type QSAR models were developed using structural, spatial, electronic, physicochemical and/or topological parameters using statistical methods like stepwise regression, partial least squares (PLS) and factor analysis followed by multiple linear regression (FA-MLR). Using topological and structural parameters, FA-MLR provided the best equation based on internal validation (Q2 = 0.514) but the best externally validated model was obtained with PLS ( = 0.565). When structural, physicochemical, spatial and electronic descriptors were used, the best Q2 value (0.562) was obtained from the stepwise regression derived model whereas the best value (0.571) came from the PLS model. When topological descriptors were used in combination with the structural, physicochemical, spatial and electronic descriptors, the best Q2 and values obtained were 0.530 (stepwise regression) and 0.580 (PLS) respectively. Attempt was made to develop 3D-QSAR models using molecular shape analysis descriptors in combination with structural, physicochemical, spatial and electronic parameters. Linear models were developed using genetic function algorithm coupled with multiple linear regression. However, the results from the 3D-QSAR study were not superior to those of the classical QSAR models. Finally, artificial neural network was employed for development of nonlinear models. The ANN models showed acceptable values of squared correlation coefficient for the observed and predicted values of the test set compounds. From the view point of external predictability, selected ANN models were superior to the linear QSAR models. All reported models satisfy the criteria of external validation as recommended by Golbraikh and Tropsha (J Mol Graphics Mod 2002; 20: 269–276), whereas the majority of the models have modified r2 () value of the test set for external validation more than 0.5 as suggested by Roy and Roy (QSAR Comb Sci 2008; 27: 302-313).

PURPOSE. Rate of brain penetration (logPS), brain/plasma equilibration rate (logPS-brain), and extent of blood-brain barrier permeation (logBB) of 29 α-adrenergic and imidazoline-receptors ligands were examined in Quantitative-Structure-Property Relationship (QSPR) study. METHODS. Experimentally determined chromatographic retention data (logKw at pH 4.4, slope (S) at pH 4.4, logKw at pH 7.4, slope (S) at pH 7.4, logKw at pH 9.1, and slope (S) at pH 9.1) and capillary electrophoresis migration parameters (μeff at pH 4.4, μeff at pH 7.4, and μeff at pH 9.1), together with calculated molecular descriptors, were used as independent variables in the QSPR study by use of partial least square (PLS) methodology. RESULTS. Predictive potential of the formed QSPR models, QSPR(logPS), QSPR(logPS-brain), QSPR(logBB), was confirmed by cross- and external validation. Hydrophilicity (Hy) and H-indices (H7m) were selected as significant parameters negatively correlated with both logPS and logPS-brain, while topological polar surface area (TPSA(NO)) was chosen as molecular descriptor negatively correlated with both logPS and logBB. The principal component analysis (PCA) and hierarchical clustering analysis (HCA) were applied to cluster examined drugs based on their chromatographic, electrophoretic and molecular properties. Significant positive correlations were obtained between the slope (S) at pH 7.4 and logBB in A/B cluster and between the logKw at pH 9.1 and logPS in C/D cluster. CONCLUSIONS. Results of the QSPR, clustering and correlation studies could be used as novel tool for evaluation of blood-brain barrier permeation of related α-adrenergic/imidazoline receptor ligands.This article is open to POST-PUBLICATION REVIEW. Registered readers (see "For Readers") may comment by clicking on ABSTRACT on the issue's contents page.PURPOSE. Rate of brain penetration (logPS), brain/plasma equilibration rate (logPS-brain), and extent of blood-brain barrier permeation (logBB) of 29 α-adrenergic and imidazoline-receptors ligands were examined in Quantitative-Structure-Property Relationship (QSPR) study. METHODS. Experimentally determined chromatographic retention data (logKw at pH 4.4, slope (S) at pH 4.4, logKw at pH 7.4, slope (S) at pH 7.4, logKw at pH 9.1, and slope (S) at pH 9.1) and capillary electrophoresis migration parameters (μeff at pH 4.4, μeff at pH 7.4, and μeff at pH 9.1), together with calculated molecular descriptors, were used as independent variables in the QSPR study by use of partial least square (PLS) methodology. RESULTS. Predictive potential of the formed QSPR models, QSPR(logPS), QSPR(logPS-brain), QSPR(logBB), was confirmed by cross- and external validation. Hydrophilicity (Hy) and H-indices (H7m) were selected as significant parameters negatively correlated with both logPS and logPS-brain, while topological polar surface area (TPSA(NO)) was chosen as molecular descriptor negatively correlated with both logPS and logBB. The principal component analysis (PCA) and hierarchical clustering analysis (HCA) were applied to cluster examined drugs based on their chromatographic, electrophoretic and molecular properties. Significant positive correlations were obtained between the slope (S) at pH 7.4 and logBB in A/B cluster and between the logKw at pH 9.1 and logPS in C/D cluster. CONCLUSIONS. Results of the QSPR, clustering and correlation studies could be used as novel tool for evaluation of blood-brain barrier permeation of related α-adrenergic/imidazoline receptor ligands. This article is open to POST-PUBLICATION REVIEW. Registered readers (see "For Readers") may comment by clicking on ABSTRACT on the issue's contents page.

Quantitative structure–property relationship (QSPR) modeling is a powerful approach for predicting environmental behavior of organic pollutants with their structure descriptors. This study reports two optimal QSPR models for estimating water solubility ( log S W ) and n-octanol/water partition coefficient ( log K OW ) of chloric and alkyl benzene derivatives. Quantum chemical descriptors computed with density functional theory at B3LYP/6-31G(d) level and partial least squares (PLS) analysis with optimizing procedure were used for generating QSPR models for log S W and log K OW of chloric and alkyl benzene derivatives. The correlation coefficients of the optimal models for log S W and log K OW were 0.973 and 0.990, respectively. The results of internal cross-validation test and external validation test showed that both of the optimal models had high fitting precision and good predicting ability. The log S W and log K OW values predicted by the optimal models are very close to those observed. The PLS analysis indicated that chloric and alkyl benzene derivatives with larger electronic spatial extent and lower molecular total energy tend to be more hydrophobic and lipophilic, and smaller energy gap between the lowest unoccupied and the highest occupied molecular orbitals leads to larger dissolvability.

In the present study, adequate quantitative structure-activity relationship (QSAR) models were developed to analyze the water solubility of some ammonium hexafluorosilicates. All models were developed using structural descriptors calculated by the SiRMS method based on the simplex representation of the molecular structure and Dragon descriptors. Log P, equalized electronegativity, molecular refraction, and molecular weight were used as integral descriptors in addition to 2D structural descriptors. All QSAR models were obtained using the partial least squares (PLS) method. Model M1 examined the influence of various physicochemical and structural factors on the water solubility of the studied compounds. The interpretation results are consistent with the qualitative data of previous experimental works. It was possible to detail the features of the hydrogen bonds effect on the water solubility for investigated compounds The non-trivial nature of the effect of the hydrogen bond was also shown. QSPR model M2 could predict the solubility of new compounds of the studied type with satisfactory accuracy.

Quantitative structure–property relationships (QSPRs) were developed using a genetic algorithm (GA), based on the variable-selection approach with topological descriptors. The selectivity coefficients of 26 molecules (drug, amino-acid and organic compound) of a histamine-selective electrode were efficiently estimated and predicted with the QSPR models. The most important descriptors were selected from a set of 74 topological descriptors to build the QSPR models, using the multiple linear regressions (MLRs) and the partial least squares (PLS) regression. The predictive quality of the QSPR models was tested for an external prediction set of 7 compounds, randomly chosen from 26 compounds. The PLS regression method was used to model the structure-selectivity coefficient relationships. However, the results surprisingly showed more or less the same quality for MLR and PLS modeling, according to the squared regression coefficients R2 values, which were 0.918 and 0.915, respectively. In addition, the theoretical investigation on the interaction of the histamine and the other studied compounds with the ionophore was performed. The correlation between the interaction energies and the selectivity coefficients of the studied compounds was equal to 0.993, demonstrating the applicability of these results for the prediction of the selectivity coefficients.

QSAR for dihydrofolate reductase inhibitors with molecular graph structural descriptors

Variable-selection approaches to generate QSAR models for a set of antichagasic semicarbazones and analogues

Mid-infrared (MIR) spectroscopy is a powerful, eco-friendly, and cost-effective technique for predicting soil property. However, its predictive accuracy can be affected by factors such as moisture content, particle size, sensor variability, and the baseline noise. To address these limitations, this study investigated the impact of combining various preprocessing techniques with variable selection methods on the performance of partial least squares regression (PLSR) models. Soil samples from the Rhamna region of Morocco were analyzed to estimate key properties, including total nitrogen (TN), total carbon (TC), total organic carbon (TOC), clay, silt, sand, moisture content (MC), pH, phosphorus (P 2 O 5 ), and cation exchange capacity (CEC). Spectral data were preprocessed using methods such as standard normal variate (SNV), Savitzky–Golay smoothing (SG smoothing), first and second derivatives (SG1D and SG2D), and their combinations (e.g., SNV + SG2D). The best-performing preprocessing combinations were then used with variable selection approaches, interval PLS (iPLS), variable importance in projection (VIP), and selectivity ratio (SR). The results indicated that Savitzky–Golay (SG) derivatives combined with SNV generally improved model performance across most soil properties. In particular, total nitrogen (TN) prediction improved primarily with the first SG derivative, with R 2 cv increasing from 0.82 (raw spectra) to 0.88 (SG1D), while RMSEcv decreased from 0.03% to 0.01%. Further improvements were achieved through variable selection, with iPLS providing the most consistent enhancement across properties with a very low number of features compared to the other methods. Overall, the integration of optimal preprocessing and iPLS variable selection significantly improved the predictive accuracy and robustness of partial least squares regression (PLSR) models for soil property estimation compared with the full spectrum.

Machine-learning-based similarity meets traditional QSAR: “q-RASAR” for the enhancement of the external predictivity and detection of prediction confidence outliers in an hERG toxicity dataset

In this paper, an efficient variable and subset selection approach related to modified partial least squares (PLS) regression are proposed and applied to numerical and industry application to show the superior of the results. The main purpose of the proposed variable selection approach is to offer further useful information to identification, control as well as diagnosis. The significant advantage of the modified PLS-based variable selection method is to keep the superior fitting and prediction performance as well as the significantly reduced number of selected variables. To this aim, the canonical and modified PLS regression methods have been introduced. Then an effective and precise variable selection method which can select the most significant elements among huge redundant information is proposed. To this end, the effectiveness of the proposed variable selection method is demonstrated by the simulation results of a numerical instance and a process industry case study.

In this paper, we compare protein- and ligand-based virtual screening techniques for identifying the ligands of four biogenic amine-binding G-protein coupled receptors (GPCRs). For the screening of the virtual compound libraries, we used (1) molecular docking into GPCR homology models, (2) ligand-based pharmacophore and Feature Tree models, (3) three-dimensional (3D)-similarity searches, and (4) statistical methods [partial least squares (PLS) and partial least squares discriminant analysis (PLS-DA) models] based on two-dimensional (2D) molecular descriptors. The comparison of the different methods in retrieving known antagonists from virtual libraries shows that in our study the ligand-based pharmacophore-, Feature Tree-, and 2D quantitative structure-activity relationship (QSAR)-based screening techniques provide enrichment factors that are higher than those provided by molecular docking into the GPCR homology models. Nevertheless, the hit rates achieved when docking with GOLD and ranking the ligands with GoldScore (up to 60% among the top-ranked 1% of the screened databases) are still satisfying. These results suggest that docking into GPCR homology models can be a useful approach for lead finding by virtual screening when either little or no information about the active ligands is available.

Quality assessment of diesel fuel is highly necessary for society, but the costs and time spent are very high while using standard methods. Therefore, this study aimed to develop an analytical method capable of simultaneously determining eight diesel quality parameters (density; flash point; total sulfur content; distillation temperatures at 10% (T10), 50% (T50), and 85% (T85) recovery; cetane index; and biodiesel content) through attenuated total reflection Fourier transform infrared (ATR-FTIR) spectroscopy and the multivariate regression method, partial least square (PLS). For this purpose, the quality parameters of 409 samples were determined using standard methods, and their spectra were acquired in ranges of 4000–650 cm−1. The use of the multivariate filters, generalized least squares weighting (GLSW) and orthogonal signal correction (OSC), was evaluated to improve the signal-to-noise ratio of the models. Likewise, four variable selection approaches were tested: manual exclusion, forward interval PLS (FiPLS), backward interval PLS (BiPLS), and genetic algorithm (GA). The multivariate filters and variables selection algorithms generated more fitted and accurate PLS models. According to the validation, the FTIR/PLS models presented accuracy comparable to the reference methods and, therefore, the proposed method can be applied in the diesel routine monitoring to significantly reduce costs and analysis time.

Quantitative structure-activity relationship (QSAR) models were developed for the in vitro potencies to downregulate gap junctional intercellular communication (GJIC) of hydroxylated polychlorinated biphenyls (OH-PCBs) and PCB quinines using partial least squares (PLS) regression. Quantum chemical descriptors computed by the semiempirical AM1, PM3 and MNDO methods were used as predictor variables. The cross-validated Q2cum values for the three optimal QSAR models are 0.784, 0.789 and 0.755, respectively, indicating good predictive capabilities for the acute inhibition of GJIC (IC(50)) of oxygenated PCB derivatives. The slightly higher Q2cum value of the model using computed molecular descriptors from the PM3 Hamiltonian suggested a slightly better predictive power than the models developed using AM1 or MNDO. However, given these dispersion parameters in these three optimal models, there would not be a significant difference between the Q2cum values. Results from this study showed that the logarithmic scale of IC(50) is affected by different molecular structural descriptors.

The pseudo first-order rate constants for reductive dehalogenation under anoxic conditions have recently been reported for a series of halogenated aliphatic hydrocarbons. In this paper it is shown that multivariate quantitative structure-activity relationship (QSAR) modelling of these data is possible. Based on a training set of nine compounds and using information from 36 chemical and biological model systems, a QSAR was developed explaining 82% (R2) and predicting 53% (Q2int) of the variation in reductive dehalogenation. The QSAR was calculated using the partial least squares (PLS) method. The derived QSAR was validated using an external validation set of six compounds for which experimentally determined rate constants were available. The external validation showed that the QSAR could predict 62% (Q2ext) of the response variation in the validation set. In addition. the validity of the QSAR was explored using a combination of cross-validation and permutation. Fifty repetitive randomizations of the real response were made and for each resulting reordered response the full PLS analysis was conducted. This yielded 50 pairs of Q2int and Q2ext which were compared with those values of the ‘real’ model. Thus it was demonstrated that the QSAR produced Q2 values significantly exceeding those of the random models and consequently the model was concluded to be valid. The interpretation of the QSAR suggested that the mechanism of reductive dehalogenation involves elements of hydrophobicity and electron transfer processes. © 1996 by John Wiley & Sons, Ltd.

QSAR by LFER model of HIV protease inhibitor mannitol derivatives using FA-MLR, PCRA, and PLS techniques