The use of Bayesian nonlinear regression techniques for the modelling of the retention behaviour of volatile components of Artemisia species.

The main aim of this work was to assess the ability of Bayesian multivariate adaptive regression splines (BMARS) and Bayesian radial basis function (BRBF) techniques for modelling the gas chromatographic retention indices of volatile components of Artemisia species. A diverse set of molecular descriptors was calculated and used as descriptor pool for modelling the retention indices. The ability of BMARS and BRBF techniques was explored for the selection of the most relevant descriptors and proper basis functions for modelling. The results revealed that BRBF technique is more reproducible than BMARS for modelling the retention indices and can be used as a method for variable selection and modelling in quantitative structure-property relationship (QSPR) studies. It is also concluded that the Markov chain Monte Carlo (MCMC) search engine, implemented in BRBF algorithm, is a suitable method for selecting the most important features from a vast number of them. The values of correlation between the calculated retention indices and the experimental ones for the training and prediction sets (0.935 and 0.902, respectively) revealed the prediction power of the BRBF model in estimating the retention index of volatile components of Artemisia species.

Classification of anti-HIV compounds using counterpropagation artificial neural networks and decision trees.

The main aim of the present work was to collect and categorize anti-HIV molecules in order to identify general structure-activity relationships. In this respect, a total of 5580 drugs and drug-like molecules was collected from 256 different articles published between 1992 and 2010. An algorithm called genetic algorithm-pattern search counterpropagation artificial neural networks (GPS-CPANN) was proposed for the classification of compounds. In addition, the CART (classification and regression trees) method was used for construction of decision trees and finding the best molecular descriptors. The results revealed that the developed CPANN models and decision tree can correctly classify the molecules according to their inhibition mechanisms and activities. Some general parameters such as molecular weight, average molecular weight, number of hydrogen atoms and number of hydroxyl groups were found to be important for describing the inhibition behaviour of anti-HIV agents. The developed classifier models in this work can be used to screen large libraries of compounds to identify those likely to display activity as anti-HIV agents.

QSAR modelling of integrin antagonists using enhanced Bayesian regularised genetic neural networks.

Bayesian regularised genetic neural network (BRGNN) has been used for modelling the inhibition activity of 141 biphenylalanine derivatives as integrin antagonists. Three local pattern search (PS) methods, simulated annealing and threshold acceptance were combined with BRGNN in the form of a hybrid genetic algorithm (HGA). The results obtained revealed that PS is a suitable method for improving the ability of BRGNN to break out from the local minima. The proposed HGA technique is able to retrieve important variables from complex systems and nonlinear search spaces for optimisation. Two models with 8-3-1 artificial neural network (ANN) architectures were developed for describing α4ß7 and α4ß1 modulatory activities of integrin antagonists. Monte Carlo cross-validation was performed to validate the models and Q2 values of 0.75 and 0.74 were obtained for α4ß7 and α4ß1 inhibitory activities, respectively. The scrambling technique was used for sensitivity analysis of descriptors appearing in ANN models. Frequencies of repetition and sensitivity analysis of molecular descriptors revealed that 3D-Morse descriptors are influential factors for describing α4ß7 inhibitory activity, while in the case of α4ß1 inhibitory activity, the Randic shape index, the lowest eigenvalue of the Burden matrix and the number of rotatable bonds are important parameters.

Shuffling multivariate adaptive regression splines and adaptive neuro-fuzzy inference system as tools for QSAR study of SARS inhibitors.

In this work, the inhibitory activity of pyridine N-oxide derivatives against human severe acute respiratory syndrome (SARS) is predicted in terms of quantitative structure-activity relationship (QSAR) models. These models were developed with the aid of multivariate adaptive regression spline (MARS) and adaptive neuro-fuzzy inference system (ANFIS) combined with shuffling cross-validation technique. A shuffling MARS algorithm is utilized to select the most important variables in QSAR modeling and then these variables were used as inputs of ANFIS to predict SARS inhibitory activities of pyridine N-oxide derivatives. A data set of 119 drug-like compounds was coded with over hundred calculated meaningful molecular descriptors. The best descriptors describing the inhibition mechanism were solvation connectivity index, length to breadth ratio, relative negative charge, harmonic oscillator of aromatic index, average molecular weight and total path count. These parameters are among topological, electronic, geometric, constitutional and aromaticity descriptors. The statistical parameters of R2 and root mean square error (RMSE) are 0.884 and 0.359, respectively. The accuracy and robustness of shuffling MARS-ANFIS model in predicting inhibition behavior of pyridine N-oxide derivatives (pIC50) was illustrated using leave-one-out and leave-multiple-out cross-validation techniques and also by Y-randomization. Comparison of the results of the proposed model with those of GA-PLS-ANFIS shows that the shuffling MARS-ANFIS model is superior and can be considered as a tool for predicting the inhibitory behavior of SARS drug-like molecules.

Quantitative structure-activity relationship study of serotonin (5-HT7) receptor inhibitors using modified ant colony algorithm and adaptive neuro-fuzzy interference system (ANFIS).

Quantitative structure-activity relationship (QSAR) approach was carried out for the prediction of inhibitory activity of some novel quinazolinone derivatives on serotonin (5-HT(7)) using modified ant colony (ACO) method and adaptive neuro-fuzzy interference system (ANFIS) combined with shuffling cross-validation technique. A modified ACO algorithm is utilized to select the most important variables in QSAR modeling and then these variables were used as inputs of ANFIS to predict 5-HT(7) receptor binding activities of quinazolinone derivatives. The best descriptors describing the inhibition mechanism are Q(max), Se, Hy, PJI3 and DELS which are among electronic, constitutional, geometric and empirical descriptors. The statistical parameters of R(2) and root mean square error are 0.775 and 0.360, respectively. The ability and robustness of modified ACO-ANFIS model in predicting inhibition behavior of quinazolinone derivatives (pIC(50)) are illustrated by validation techniques of leave-one-out and leave-multiple-out cross-validations and also by Y-randomization technique. Comparison of the modified ACO-ANFIS method with two other methods, that is, stepwise MLR-ANFIS and GA-PLS-ANFIS were also studied and the results indicated that the proposed model in this work is superior over the others.

Model-based analysis for kinetic complexation study of Pizda and Cu(II).

In the present work, the multivariate kinetic complexation of a new synthesized ligand, 1-(2''-hydroxyl cyclohexyl)-3'-[aminopropyl]-4-[3'-aminopropyl]piperazine (Pizda) and Cu(2+) in 50% ethanol-water solution is investigated using the UV-vis stopped-flow technique and state-of-the-art multi-wavelength numerical analysis. Model-based least squares fitting analysis or hard modeling is a specific part of chemometrics which is based on mathematical relationships for describing the measurements. Some recent developments include the incorporation of the effects of non-ideal experimental conditions into the fitting algorithm so it can substantially simplify experimental procedures. In this study no buffers are required because pH changes are taken into computations. Some 21 multi-wavelength kinetic measurements, taken at various initial concentrations of [H(+)] were analyzed globally, i.e. simultaneously applying an all inclusive reaction mechanism and a common set of species spectra. Using numerical analysis, the pH of the experimental solutions was allowed to vary as a consequence of the proceeding reactions. This enabled the complete kinetic analysis of the formation and dissociation of Cu(Pizda)(n+). Here protonation equilibria have been directly incorporated into the rate law, so thus variable pH values have been allowed during each measurement. Using the independently estimated stability constants (from spectrophotometric and potentiometric measurements) for the Cu(Pizda)(n+) complexes, a total of six rate constants and one protonation constant could be elucidated. The results of the analysis include the concentration distribution and spectra of all chemical species involved in the reaction. A low standard deviation and residual profiles obtained validate the results.

Comparison between two different hemichromes of hemoglobins (HbA and HbS) induced by n-dodecyl trimethylammonium bromide: chemometric study.

The interaction of n-dodecyl trimethylammonium bromide (DTAB) with oxyhemoglobin A and oxyhemoglobin S is investigated using UV-visible absorption spectra and chemometric resolution techniques. Oxyhemoglobins (A and S) induced to partial oxidized form (ferrihemoglobin) by DTAB and finally transform to fully oxidized hemichrome. Hemichrome mole fractions of HbS are more than HbA because of more hydrophobic interaction of DTAB-HbS in second set of binding site relative to DTAB-HbA. The visible spectra between 500 and 650 nm are used for identifying the present components in solution because each species of hemoglobin has a specific spectrum in this region. The number of components and mole fraction of mentioned species were determined by employing chemometric resolution techniques. Subspace comparison was used for determination of the number of components in each concentration of hemoglobin and DTAB. After the determination of components, multivariate curve resolution-alternating least square (MCR-ALS) by initial estimates of spectral profiles and proper constraints, was used to resolve the data matrix into pure concentration and spectral profiles. The results show that both number and mole fraction of components which were formed during hemoglobin (HbA and HbS) oxidation by DTAB were initial hemoglobin concentrations independent. Furthermore, in average the mole fraction of hemichrome of HbS is 14.4% more than HbA. On the other hand, the mole fraction of HbA ferrihemoglobin is 15.6% higher than HbS averagely.

QSAR study of heparanase inhibitors activity using artificial neural networks and Levenberg-Marquardt algorithm.

A linear and non-linear quantitative structure-activity relationship (QSAR) study is presented for modeling and predicting heparanase inhibitors' activity. A data set that consisted of 92 derivatives of 2,3-dihydro-1,3-dioxo-1H-isoindole-5-carboxylic acid, furanyl-1,3-thiazol-2-yl and benzoxazol-5-yl acetic acids is used in this study. Among a large number of descriptors, four parameters classified as physico-chemical, topological and electronic indices are chosen using stepwise multiple regression technique. The artificial neural networks (ANNs) model shows superiority over the multiple linear regressions (MLR) by accounting 87.9% of the variances of antiviral potency of the heparanase inhibitors. This paper focuses on investigating the role of weight update functions in developing ANNs. Levenberg-Marquardt (L-M) algorithm shows a better performance compared with basic back propagation (BBP) and conjugate gradient (CG) algorithms. The accuracy of 4-3-1 L-M ANN model was illustrated using leave-one-out (LOO), leave-multiple-out (LMO) cross-validations and Y-randomization. The mean effect of descriptors and sensitivity analysis show that log P is the most important parameter affecting the inhibitory behavior of the molecules.

Determination of essential oil components of Artemisia haussknechtii Boiss. using simultaneous hydrodistillation-static headspace liquid phase microextraction-gas chromatography mass spectrometry.

A novel method for extraction and analysis of volatile compounds of Artemisia haussknechtii Boiss., using simultaneous hydro-distillation and static headspace liquid microextraction followed by gas chromatography-mass spectrometry (SHD-SHLPME-GCMS) is developed. SHLPME parameters including nature of extracting solvent, headspace volume and design, extraction time, sample weight and microdrop volume were optimized. Comparison of hydro-distillation gas chromatography-mass spectrometry and HD-SHLPME-GCMS showed that the latter method is fast, simple, inexpensive and effective for the analysis of volatile compounds of aromatic plants. By using this method, 56 compounds were extracted and identified for Artemisia haussknechtii Boiss. The main constituents of its essential oil that were extracted by HD-SHLPME method, include camphor (41.01%), 1,8-cineole (32.35%), cis-davanone (3.68%), 4-terpineol (2.99%), linalool (2.84%), beta-fenchyl alcohol (2.72%), and borneol (2.58%).

Optimization of solid-phase microextraction of volatile phenols in water by a polyaniline-coated Pt-fiber using experimental design.

Solid-phase microextraction (SPME) coupled to gas chromatography (GC) was applied to the extraction of phenol and some of its volatile derivatives in water samples. The SPME fiber consisted of a thin layer of polyaniline, which was electrochemically coated on a fine Pt wire. The stability of the coating was such that it could be used at temperatures as high as 325 degrees C, without any deterioration. The effects of various parameters affecting the extraction efficiency were studied, simultaneously. From these, optimization of the extraction temperature, extraction time, coating thickness, sample pH, salt concentration and desorption time was carried out by means of a (2(6-2)) fractional factorial design. It was found that the effects and interactions of five out of six factors were significant. However, the coating thickness showed a large main effect but an insignificant interaction effect, so it was kept constant. Also, the effect of desorption time was insignificant if sufficient time was allowed for desorption to take place. Therefore, a central composite design (CCD) with four remaining factors, i.e., sample pH, salt concentration, extraction time and sample temperature was performed and a response surface equation was derived. The statistical parameters of the derived model were r=0.97 and F=25.3. The optimum conditions were obtained using a grid method. Using the optimum conditions, the method was analytically evaluated. The detection limit, relative standard deviation, linear range and recovery were 1.3-12.8 ng mL(-1), 2.2-5.3%, 0.01-5.0 microg mL(-1), and 88-103%, respectively. The results showed the suitability of polyaniline-coated fiber in analyzing volatile phenolic compounds in water samples.

Simulation of 13C nuclear magnetic resonance spectra of lignin compounds using principal component analysis and artificial neural networks.

Theoretical models relating atom-based structural descriptors to 13C NMR chemical shifts were used to accurately simulate 13C NMR spectra of lignin model compounds (poly-substituted phenols). The structure-activity relationship (SAR) studies for 15 lignins using pattern recognition methods of principal component analysis (PCA) and artificial neural networks (ANNs) were performed in this work. The most important parameters affecting the 13C chemical shifts of different carbons were descriptors consisting of the charge density of the atoms at different distances from the center carbon. Among the large number of parameters, these descriptors were selected using PCA and were used as ANN input. The least square regression analyses of the results indicate correlation coefficient (R) values in excess of 0.983 for the total data set.

Use of computer-assisted methods for the modeling of the retention time of a variety of volatile organic compounds: a PCA-MLR-ANN approach.

A hybrid method consisting of principal component analysis (PCA), multiple linear regressions (MLR), and artificial neural network (ANN) was developed to predict the retention time of 149 C(3)-C(12) volatile organic compounds for a DB-1 stationary phase. PCA and MLR methods were used as feature-selection tools, and a neural network was employed for predicting the retention times. The regression method was also used as a calibration model for calculating the retention time of VOCs and investigating their linear characteristics. The descriptors of the total information index of atomic composition, IAC, Wiener number, W, solvation connectivity index, X1sol, and number of substituted aromatic C(sp(2)), nCaR, appeared in the MLR model and were used as inputs for the ANN generation. Appearance of these parameters shows the importance of the dispersion interactions in the mechanism of retention. Comparison of the MLR and 5-2-1 ANN models indicates the superiority of the ANN over that of the MLR model. The values of 0.913 and 0.738 were obtained for the standard error of prediction set of MLR and ANN models, respectively.

Prediction of gas chromatographic retention indices of a diverse set of toxicologically relevant compounds.

For a set of 846 organic compounds, relevant in forensic analytical chemistry, with highly diverse chemical structures, the gas chromatographic Kovats retention indices have been quantitatively modeled by using a large set of molecular descriptors generated by software Dragon. Best and very similar performances for prediction have been obtained by a partial least squares regression (PLS) model using all considered 529 descriptors, and a multiple linear regression (MLR) model using only 15 descriptors obtained by a stepwise feature selection. The standard deviations of the prediction errors (SEP), were estimated in four experiments with differently distributed training and prediction sets. For the best models SEP is about 80 retention index units, corresponding to 2.1-7.2% of the covered retention index interval of 1110-3870. The molecular properties known to be relevant for GC retention data, such as molecular size, branching and polar functional groups are well covered by the selected 15 descriptors. The developed models support the identification of substances in forensic analytical work by GC-MS in cases the retention data for candidate structures are not available.

Prediction of relative response factors of electron-capture detection for some polychlorinated biphenyls using chemometrics.

The relative response factor (RRF) of an electron-capture detection (ECD) system is predicted for a set of 118 polychlorinated biphenyls (PCBs). Due to the wide range of relative retention times of PCB congeners, the RRFs of these compounds were calculated based on two different internal standards. Therefore, the compounds were divided into two molecular subsets. As a first step, multiple linear regression (MLR) was employed to find informative descriptors that can predict the RRFs of these compounds. Two descriptors of molecular ion ionization potential (MIIP) and ionization potential of the molecule (IP) that are related to affinity of the compounds for the electrons show the highest mean effects in subsets 1 and 2. respectively. The descriptors appearing in the MLR models were considered as inputs for developing the back-propagation artificial neural networks (BP-ANN). Two networks with the architectures of 5-5-1 and 7-6-1 were generated for the prediction of RRFs of molecules of subsets 1 and 2, respectively. Comparison of the results indicates the superiority of neural networks over that of the MLR method indicating the nonlinear behaviors of the ECD system. Inspection of the models reveals that the surface of the molecules play different roles in response factors of two subsets due to rotation of one phenyl group with respect to the other for the subset consisting of larger number of chlorine atoms.

Prediction of electrophoretic mobilities of alkyl- and alkenylpyridines in capillary electrophoresis using artificial neural networks.

The electrophoretic mobilities of 31 isomeric alkyl- and alkenylpyridines in capillary electrophoresis were predicted using an artificial neural network (ANN). The multiple linear regression (MLR) technique was used to select the descriptors as inputs for the artificial neural network. The neural network is a fully connected back-propagation model with a 3-6-1 architecture. The results obtained using the neural network were compared with those obtained using the MLR technique. Standard error of training and standard error of prediction were 6.28 and 5.11%, respectively, for the MLR model and 1.03 and 1.20%, respectively, for the ANN model. Two geometric parameters and one electronic descriptor that were used as inputs in the ANN are able to distinguish between the isomers.

Prediction of relative response factors for flame ionization and photoionization detection using self-training artificial neural networks.

The relative response factors (RRFs) of a flame ionization detection (FID) system and two pulsed discharge photoionization detection (PID) systems with different discharge gases are predicted for a set of organic compounds containing various functional groups. As a first step, numerical descriptors were calculated based on the molecular structures of compounds. Then, multiple linear regression (MLR) was employed to find informative subsets of descriptors that can predict the RRFs of these compounds. The selected MLR model for the FID system includes seven descriptors and two selected MLR models for the PID systems with argon- and krypton-doped helium as the discharge gases, respectively, include six and five descriptors. The descriptors appearing in the MLR models were considered as inputs for the self-training artificial neural networks (STANNs). A 7-7-1 STANN was generated for prediction of RRFs of the FID system, and two STANNs with the topologies of 6-7-1 and 5-6-1 were generated for the two PID systems. Comparison of the results indicates the superiority of neural networks over that of the MLR method. This is due to the nonlinear behaviors of relative response factors for all type of detectors studied in this work.

Use of self-training artificial neural networks in modeling of gas chromatographic relative retention times of a variety of organic compounds.

A quantitative structure-activity relationship study based on multiple linear regression (MLR), artificial neural network (ANN), and self-training artificial neural network (STANN) techniques was carried out for the prediction of gas chromatographic relative retention times of 13 different classes of organic compounds. The five descriptors appearing in the selected MLR model are molecular density, Winer number, boiling point, polarizability and square of polarizability. A 5-6-1 ANN and a 5-4-1 STANN were generated using the five descriptors appearing in the MLR model as inputs. Comparison of the standard errors and correlation coefficients shows the superiority of ANN and STANN over the MLR model. This is due to the fact that the retention behaviors of molecules show non-linear characteristics. Inspection of the results of STANN and ANN shows there are few differences between these methods. However, optimization of STANN is much faster and the number of adjustable parameters for this technique is much less compared with those of the conventional ANN.

Prediction of electrophoretic mobilities of sulfonamides in capillary zone electrophoresis using artificial neural networks.

Artificial neural networks (ANNs) were successfully developed for the modeling and prediction of electrophoretic mobility of a series of sulfonamides in capillary zone electrophoresis. The cross-validation method was used to evaluate the prediction ability of the generated networks. The mobility of sulfonamides as positively charged species at low pH and negatively charged species at high pH was investigated. The results obtained using neural networks were compared with the experimental values as well as with those obtained using the multiple linear regression (MLR) technique. Comparison of the results shows the superiority of the neural network models over the regression models.

Artificial neural network modeling of Kováts retention indices for noncyclic and monocyclic terpenes.

A quantitative structure-property relationship study based on multiple linear regression (MLR) and artificial neural network (ANN) techniques was carried out to investigate the retention behavior of some terpenes on the polar stationary phase (Carbowax 20 M). A collection of 53 noncyclic and monocyclic terpenes was chosen as data set that was randomly divided into two groups, a training set and a prediction set consist of 41 and 12 molecules, respectively. A total of six descriptors appearing in the MLR model consist of one electronic, two geometric, two topological and one physicochemical descriptors. Except for the geometric parameters the remaining descriptors have a pronounced effect on the retention behavior of the terpenes. A 6-5-1 ANN was generated by using the six descriptors appearing in the MLR model as inputs. The mean of relative errors between the ANN calculated and the experimental values of the Kováts retention indexs for the prediction set was 1.88%. This is in aggreement with the relative error obtained by experiment.

Prediction of thermal conductivity detection response factors using an artificial neural network.

The main aim of the present work was the development of a quantitative structure-activity relationship method using an artificial neural network (ANN) for predicting the thermal conductivity detector response factor. As a first step a multiple linear regression (MLR) model was developed and the descriptors appearing in this model were considered as inputs for the ANN. The descriptors of molecular mass, number of vibrational modes of the molecule, molecular surface area and Balaban index appeared in the MLR model. In agreement with the molecular diameter approach, molecular mass and molecular surface area play a major role in estimating the thermal conductivity detector response factor (TCD-RF). A 4-7-1 neural network was generated for the prediction of the TCD-RFs of a collection of 110 organic compounds including hydrocarbons, benzene derivatives, esters, alcohols, aldehydes, ketones and heterocyclics. The mean absolute error between the ANN calculated and the experimental values of the response factors was 0.02 for the prediction set.