Application of GA-PLS and GA-KPLS calculations for the prediction of the retention indices of essential oils

Abstract

Genetic algorithm and partial least square (GA-PLS) and kernel PLS (GA-KPLS) techniques were used to investigate the correlation between retention indices (RI) and descriptors for 117 diverse compounds in essential oils from 5 Pimpinella species gathered from central Turkey which were obtained by gas chromatography and gas chromatography-mass spectrometry. The square correlation coefficient leave-group-out cross validation (LGO-CV) (Q²) between experimental and predicted RI for training set by GA-PLS and GA-KPLS was 0.940 and 0.963, respectively. This indicates that GA-KPLS can be used as an alternative modeling tool for quantitative structure-retention relationship (QSRR) studies.

essential oils; gas chromatography-mass spectrometry; QSRR

ARTIGO

Application of GA-PLS and GA-KPLS calculations for the prediction of the retention indices of essential oils

Hadi NoorizadehI, ^*; Abbas Farmany^I; Mehrab Noorizadeh^II

^IDepartment of Chemistry, Ilam Branch, Islamic Azad University, Ilam, Iran

^IIYoung Researchers Club, Ilam Branch, Islamic Azad University, Ilam, Iran

ABSTRACT

Genetic algorithm and partial least square (GA-PLS) and kernel PLS (GA-KPLS) techniques were used to investigate the correlation between retention indices (RI) and descriptors for 117 diverse compounds in essential oils from 5 Pimpinella species gathered from central Turkey which were obtained by gas chromatography and gas chromatography-mass spectrometry. The square correlation coefficient leave-group-out cross validation (LGO-CV) (Q²) between experimental and predicted RI for training set by GA-PLS and GA-KPLS was 0.940 and 0.963, respectively. This indicates that GA-KPLS can be used as an alternative modeling tool for quantitative structure-retention relationship (QSRR) studies.

Keywords: essential oils; gas chromatography-mass spectrometry; QSRR.

INTRODUCTION

Essential oils also called volatile or ethereal oils are aromatic oily liquids obtained from different plant parts and widely used as food flavours.¹ The composition of essential oil has been extensively investigated because of its commercial interest in the fragrance industry (soaps, colognes, perfumes, skin lotion and other cosmetics), in aromatherapy (relaxant), in cancer treatment, in pharmaceutical preparations for its therapeutic effects as a sedative, spasmolytic, antioxidant, antiviral and antibacterial agent. Recently it has also been employed in food manufacturing as natural flavouring for beverages, ice cream, candy, baked goods and chewing gum.^2,3Pimpinella is representing in Turkey by 23 spp. (5 endemic), 2 subspecies, and 2 varieties representing 27 taxa.⁴ The most widely known and cultivated Pimpinella species is P. anisum. Pimpinella anisum (Anis) fruits (Aniseed) have been use in Turkish folk medicine as carminative, appetizers, sedative, and agents to increase milk secretion.⁵Aniseed is an important agricultural crop of Turkey. The main volatile compounds found in Pimpinella oils were monoterpenes, sesquiterpenes, trinorsesquiterpenes and phenylpropanoids (propenylphenols, pseudoisoeugenols). These entire compounds have been identified by gas chromatography and gas chromatography-mass spectrometry (GC-MS).

However, in the case of GC the identification of chromatographic peaks is just carry out by means of comparison of retention indices with reference compounds. This means that due to the complexity of the matrix, some of the components may not be identify. On the other hand, such a study requires the availability of the standards for all components of the essential oil that sometimes are not available. For GC-MS technique, much more components are qualitatively and quantitatively analyzed, but the determination is performed only through the direct similarity searches in MS database attached to the GC-MS instruments. There exist at least two serious problems for this approach. First, the background cannot be accurately corrected. Second, there are always overlapping/ embedded peaks even under good separating conditions. Both problems can possibly result in wrong similarity matches in the MS library. In recent decades, by developing hyphenated techniques such as GC-MS and high performed liquid chromatography-mass spectrometry (HPLC-MS), with two-dimensional data, which can provide information both in the chromatographic and spectral directions in one run, have become available. Thus, the qualitative and quantitative analysis of components can be performed not only with retention indices, peak heights and areas but also by ultra violent (UV) and/or mass spectra.^6,7

Chromatographic retention for capillary column gas chromatography is the calculated quantity, which represents the interaction between stationary liquid phase and gas-phase solute molecule. This interaction can be related to the functional group, electronic and geometrical properties of the molecule.^8,9

Mathematical modeling of these interactions helps chemists to find a model that can be used to obtain a deep understanding about the mechanism of interaction and to predict the retention indices (RI) of new or even unsynthesized compounds.¹⁰ Building retention prediction models may initiate such theoretical approach, and several possibilities for retention prediction in GC. Among all methods, quantitative structure-retention relationships (QSRR) are most popular. In QSRR, the retention of given chromatographic system was modeled as a function of solute (molecular) descriptors. A number of reports, deals with QSRR retention indices calculation of several compounds have been published in the literature.^11-13 The QSRR models apply to partial least squares(PLS) method often combined with genetic algorithms (GA) for feature selection.^14,15

Because of the complexity of relationships between the property of molecules and structures, nonlinear models are also used to model the structure-property relationships. Nonlinear kernel-based algorithms as kernel partial least squares (KPLS) have been proposed.^{16, 17} The basic idea of KPLS is first to map each point in an original data space into a feature space via nonlinear mapping and then to develop a linear PLS model in the mapped space. According to Cover's theorem, nonlinear data structure in the original space is most likely to be linear after high-dimensional nonlinear mapping.¹⁸ Therefore, KPLS can efficiently compute latent variables in the feature space by means of integral operators and nonlinear kernel functions. Compared to other nonlinear methods, the main advantage of the kernel based algorithm is that it does not involve nonlinear optimization. It essentially requires only linear algebra, making it as simple as the conventional linear PLS. In addition, because of its ability to use different kernel functions, KPLS can handle a wide range of nonlinearities. In the present study, GA-PLS and GA-KPLS were employed to generate QSRR models that correlate the structure of some compound; with observed RI.

EXPERIMENTAL

Data set

Retention indices of essential oils from 5 Pimpinella species gathered from central Turkey (Pimpinella anisetum, Pimpinella anisum, Pimpinella cappadocica var. cappadocica, Pimpinella flabellifolia and Pimpinella isaurica) was studied by GC-FID and GC-MS, which contains 117 compounds¹⁹ (Table 1). Quantification of essential oil components was performed on the basis of their GC-FID peak areas on the Innowax column and percentages of the characterized components. The identification of the volatile organic compounds was achieved through retention indices and mass spectrometry by comparison mass spectra of the unknown peaks with those stored in the Wiley GC-MS Library, MassFinder and the in-house "Baser Library of Essential Oil Constituents" which includes over 3200 genuine compounds with MS and retention data. The n-Alkanes (C9-C20) were used as reference points in the calculation of retention indices. Essential oils of Pimpinella species were subjected to silica gel column chromatography using n-hexane and diethyl ether according to previous procedures.¹⁹ Structure elucidation of isolated compounds was achieved by 1D and 2D NMR techniques (Bruker Avance DRX 300, 400 and 500) and LC-electrospray ionization-MS and GC-MS were used to confirm molecular weights. Isolated compounds were re-analyzed by GC-MS to confirm their identity with Pimpinella essential oil constituents and mass spectral fragmentation patterns. Essential oils were analyzed by GC using a Agilent Technologies 6890 system (SEM, Istanbul, Turkey) and an HP Innowax FSC (60 m × 0.25 mm i.d., 0.25 μm film thickness) with nitrogen as the carrier gas at 1 mL/min. Flame ionization detection and injector temperatures were performed at 250 ºC. GC-MS analysis was also performed by Hewlett Packard G1800A GCD system (SEM, Istanbul, Turkey) and an HP Innowax FSC column (60 m × 0.25 mm, 0.25 μm film thickness) was used with helium as the carrier gas (0.7 mL/min). GC oven temperature and analytical conditions were as described above. Mass spectra were recorded at 70 eV. Mass range was from m/z 35 to 425. The RI of these compounds was decreased in the range of 2931 and 1032 for both Hexadecanoic acid and α-Pinene, respectively.

In order to evaluate the generated models, we used leave-group-out cross validation (LGO-CV). This methodology systematically removed one group data at a time from the data set. A QSRR model was then constructed on the basis of this reduced data set and subsequently used to predict the removed data set. This procedure was repeated until a complete set of predicted was obtained.

Descriptor calculation

All structures were drawn with the HyperChem software (version 6). Optimization of molecular structures was carried out by semi-empirical AM1 method using the Fletcher- Reeves algorithm until the root mean square gradient of 0.01 was obtained. Since the calculated values of the electronic features of molecules will be influenced by the related conformation. In the current research an attempt was made to use the most stable conformations. Some electronic descriptors such as dipole moment and orbital energies of LUMO and HOMO were calculated by using the HyperChem software. Also optimized structures were used to calculate 1497 descriptors by DRAGON software²⁰ version 3.

Genetic algorithm

Genetic algorithm has been proposed by J. Holland in the early 1970s but it was possible to apply them with reasonable computing times only in the 1990s, when computers became much faster. GA is a stochastic method to solve the optimization problems, defined by fitness criteria applying to the evolution hypothesis of Darwin and different genetic functions, i.e., crossover and mutation.²¹ Compared to the traditional search and optimization procedures, GA is robust, global and generally more straightforward to apply to situations where there is little or no a priori knowledge about the process to be controlled. Since GA does not require derivative information or a formal initial estimate of the solution region and because of the stochastic nature of the search mechanism, it is capable to search the entire solution space with a greater probability of finding the global optimum.²² In GA, each individual of the population, defined by a chromosome of binary values as the coding technique, represented a subset of descriptors. The number of genes at each chromosome was equal to the number of descriptors. The population of the first generation was selected randomly. A gene was given the value of one, if its corresponding descriptor was included in the subset; otherwise, it was given the value of zero. The GA performs its optimization by variation and selection via the evaluation of the fitness function η. Fitness function was used to evaluate alternative descriptor subsets that were finally ordered according to the predictive performance of related model by cross validation. The fitness function was proposed by Depczynski et al..²³ The root-mean-square errors of calibration (RMSEC) and prediction (RMSEP) were calculated and the fitness function was calculated by Equation 1

where m_c and m_p are the number of compounds in the calibration and prediction set and n represent the number of selected variables, respectively. The parameter algorithm reported in Table 2.

Linear model

Partial least squares

PLS is a linear multivariate method for relating the process variables X with responses Y. PLS can analyze data with strongly collinear, noisy, and numerous variables²⁴ in both X and Y. PLS reduces the dimension of the predictor variables by extracting factors or latent variables that are correlated with Y while capturing a large amount of the variations in X. This means that PLS maximizes the covariance between matrices X and Y. In PLS, the scaled matrices X and Y are decomposed into score vectors (t and u), loading vectors (p and q), and residual error matrices (E and F):

where a is the number of latent variables. In an inner relation, the score vector t is linearly regressed against the score vector u.

where b is regression coefficient that is determined by minimizing the residual h. It is crucial to determine the optimal number of latent variables and cross validation is a practical and reliable way to test the predictive significance of each PLS component. There are several algorithms to calculate the PLS model parameters. In this work, the NIPALS algorithm was used with the exchange of scores.²⁵

NONLINEAR MODEL

Kernel partial least squares

The KPLS method is based on the mapping of the original input data into a high dimensional feature space ℑ where a linear PLS model is created. By nonlinear mapping Φ:x∈ℜⁿ → Φ(x)∈ℑ, a KPLS algorithm can be derived from a sequence of NIPALS steps and has the following formulation:²⁶ 1. Initialize score vector w as equal to any column of Y. 2. Calculate scores u = ΦΦ^Tw and normalize u to ||u|| = 1, where Φ is a matrix of regressors. 3. Regress columns of Y on u: c = Y^Tu, where c is a weight vector. 4. Calculate a new score vector w for Y: w = Yc and then normalize w to ||w||=1. 5. Repeat steps 2-4 until convergence of w. 6. Deflate ΦΦ^T and Y matrices:

7. Go to step 1 to calculate the next latent variable.

Without explicitly mapping into the high-dimensional feature space, a kernel function can be used to compute the dot products as follows:

ΦΦ^T represents the (n×n) kernel Gram matrix K of the cross dot products between all mapped input data points Φ(x_i),i = 1,…,n. The deflation of the ΦΦ^T = K matrix after extraction of the u components is given by:

where I is an m-dimensional identity matrix. Taking into account the normalized scores u of the prediction of KPLS model on training data Ŷ is defined as:

For predictions on new observation data Ŷ_t, the regression can be written as:

where K_t is the test matrix whose elements are K_ij =K(x_i, x_j) where x_i and x_j present the test and training data points, respectively.

Software and programs

A Pentium IV personal computer (CPU at 3.06 GHz) with windows XP operational system was used. Geometry optimization was performed by HyperChem (version 7.0 Hypercube, Inc.); Dragon software was used to calculate of descriptors. MINITAB software (version 14, MINITAB) was used for the simple PLS analysis. Cross validation, GA-PLS, GA-KPLS and other calculation were performed in the MATLAB (Version 7, Mathworks, Inc.) environment.

Model validation

Validation is a crucial aspect of any QSPR/QSRR modeling.²⁷ The accuracy of proposed models was illustrated using the evaluation techniques such as leave-group-out cross validation (LGO-CV) procedure and validation through an external test set.

Cross validation technique

Cross validation is a popular technique used to explore the reliability of statistical models. Based on this technique, a number of modified data sets are created by deleting in each case one or a small group (leave-some-out) of objects. For each data set, an input-output model is developed, based on the utilized modeling technique. Each model is evaluated, by measuring its accuracy in predicting the responses of the remaining data (the ones or group data that have not been utilized in the development of the model).²⁸ In particular, the LGO procedure was utilized in this study. A QSRR model was then constructed on the basis of this reduced data set and subsequently used to predict the removed data. This procedure was repeated until a complete set of predicted was obtained. The statistical significance of the screened model was judged by the correlation coefficient (Q²). The predictive ability was evaluated by the cross validation coefficient (Q² or R²_cv) which is based on the prediction error sum of squares (PRESS) and was calculated by following equation:

Where y_i, ŷ_i and y^- were respectively the experimental, predicted, and mean RI values of the samples. The accuracy of cross validation results is extensively accepted in the literature considering the Q² value. In this sense, a high value of the statistical characteristic (Q² > 0.5) is considered as proof of the high predictive ability of the model.²⁹ However, this assumption is in many cases incorrect and can be that exist the lack of the correlation between the high Q² and the high predictive ability of QSPR/QSRR models has been established and corroborated recently.²⁷ Thus, the high value of Q² appears to be necessary but not sufficient condition for the models to have a high predictive power. These authors stated that an external set is necessary. As a next step, further analysis was also followed for chemical property of the new set of compounds using the developed QSRR model.

Validation through the external test set

Validating QSRR with external data (i.e. data not used in the model development) is the best method of validation. However the availability of an independent external test set of several compounds is rare in QSRR. Thus, the predictive ability of a QSRR model with the selected descriptors was further explored by dividing the full data set. The predictive power of the models developed on the selected training set is estimated on the predicted values of test set chemicals. The data set was randomly divided into two groups including training set (calibration and prediction sets) and test set, which consists of 92 and 25 molecules, respectively. The calibration set was used for model generation. The prediction set was applied deal with overfitting of the network, whereas test set which its molecules have no role in model building was used for the evaluation of the predictive ability of the models for external set. The result clearly displays a significant improvement of the QSRR model consequent to non-linear statistical treatment and a substantial independence of model prediction from the structure of the test molecule. In the above analysis, the descriptive power of a given model has been measured by its ability to predict partition of unknown drugs. For instance, as to prediction ability, it can be observed in Figure 1 that scattering of data points from the ideal trend in test set is poor.

RESULTS AND DISCUSSION

Results of linear model

To reduce the original pool of descriptors to an appropriate size, the objective descriptor reduction was performed using various criteria. Reducing the pool of descriptors eliminates those descriptors which contribute with no information or whose information content is redundant with other descriptors present in the pool. After this process, 1007 descriptors remained. These descriptors were employed to generate the models with the GA-PLS and GA KPLS program. The best model is selected on the basis of the highest multiple correlation coefficient leave-group-out cross validation (LGO-CV) (Q²), as well as the simplicity of the model. The best PLS model contains 10 selected descriptors in 4 latent variables space. These descriptors were obtained constitutional descriptors (number of atoms (nAT), number of bonds (nBT) and number of Hydrogen atoms (nH)), topological descriptors (polarity number (Pol)) and (Balaban distance connectivity index (J)), Burden eigenvalues (lowest eigenvalue n. 2 of Burden matrix/weighted by atomic polarizabilities (BELp2) and lowest eigenvalue n. 6 of Burden matrix/weighted by atomic polarizabilities (BELp6)), atom-centred fragments (CH3R/CH4 (C-001)) and electronic descriptor (lowest unoccupied molecular orbital (LUMO) and dipole moment (µ)). For this in general, the number of components (latent variables) is less than number of independent variables in PLS analysis. The obtained statistic parameters of the GA-PLS model were shown in Tables 3 and 4. The PLS model uses higher number of descriptors that allow the model to extract better structural information from descriptors to result in a lower prediction error. The statistical parameters R², Q², least root mean squares error (RMSE), absolute error (AbsE) and relative error (RE) of prediction were obtained for GA-PLS proposed model. These parameters are probably the most popular measure of how well a model fits the data. The Root Mean Square Error (RMSE) is a frequently used measure of the difference between values predicted by a model and the values experimentally. The RMSE of a model prediction with respect to the estimated variable X_model is defined as the square root of the mean squared error:

where X_obs is observed values and X_model is modeled values at time/place i. The PLS model uses a higher number of descriptors that allow the model to extract better structural information from descriptors in order to result in a lower prediction error. With this mathematical model, we do not intend to explain the chromatographic phenomena but we propose a tool for the prediction of the performance of non-evaluated compounds.

Results of nonlinear model

With the aim of improving the predictive performance of QSRR model, GA-KPLS modeling was performed. The leave-group-out cross validation (LGO-CV) has been performed. The n selected descriptors in each chromosome were evaluated by fitness function of PLS and KPLS based on the Equation 1. In this paper a radial basis kernel function, k(x,y)= exp(||x-y||²/c), was selected as the kernel function with c = rmΣ² where r is a constant that can be determined by considering the process to be predicted (here r set to be 1), m is the dimension of the input space and Σ² is the variance of the data.³⁰ It means that the value of c depends on the system under the study. The 7 descriptors in 2 latent variables space chosen by GA-KPLS feature selection methods were contained constitutional descriptors (number of atoms (nAT), mean atomic Sanderson electronegativity (scaled on Carbon atom) (Me)), Burden eigenvalues (lowest eigenvalue n. 2 of Burden matrix/weighted by atomic polarizabilities (BELp2), highest eigenvalue n. 2 of Burden matrix/weighted by atomic van der Waals volumes (BEHv2)), charge descriptors (maximum negative charge (qnmax)) and electronic descriptors (dipole moment (µ) and high occupied molecular orbital (HOMO)).

For the constructed model, 5 general statistical parameters were selected to evaluate the prediction ability of the model for the RI. The statistical parameters R², Q², RE, AbsE and RMSE were obtained for proposed models. Each of the statistical parameters mentioned above were used for assessing the statistical significance of the QSRR model. Inspection of the results reveals a higher R² and lowers other values parameter for the training and test sets GA-KPLS compared with their counterparts for GA-PLS. The GA-PLS linear model has good statistical quality with low relative error, while the corresponding errors obtained by the GA-KPLS model are lower. Plots of predicted RI by GA-KPLS versus experimental RI values for training and test set are shown Figure 1. Obviously, there is a close agreement between the experimental and predicted RI and the data represent a very low scattering around a straight line with respective slope and intercept close to one and zero. This clearly shows the strength of GA-KPLS as a nonlinear feature selection method. This result indicates that the RI of essential oils possesses some nonlinear characteristics.

Description of models descriptors

In the chromatographic retention of compounds in the polar stationary phases is related to the induced forces that are very important in retention of the compounds. The induced forces are related to the dipolar moment, which should stimulate dipole-induced dipole interactions. Also, it is related to the dispersion forces. The dispersion forces related to steric factors, molecular size and branching. For these reasons, constitutional descriptors and electronic descriptors are very important.

Electronic descriptors were defined in terms of atomic charges and used to describe electronic aspects both of the whole molecule and of particular regions, such atoms, bonds, and molecular fragments. This descriptor calculated by computational chemistry and therefore can be consider among quantum chemical descriptor. The eigenvalues of LUMO and HOMO and their energy gap reflect the chemical activity of the molecule. LUMO as an electron acceptor represents the ability to obtain an electron, while HOMO as an electron donor represents the ability to donate an electron. HOMO energy is a useful descriptor that presents information on the distribution of π electron and explains π-π charge transfer interactions of unsaturated compounds. The lowest unoccupied molecular orbital (LUMO) energy can be interpreted as a measure of charge transfer interactions and/or of hydrogen bonding effects.^31,32

Constitutional descriptors are most simple and commonly used descriptors, reflecting the molecular composition of a compound without any information about its molecular geometry. The most common Constitutional descriptors are number of atoms, number of bound, absolute and relative numbers of specific atom type, absolute and relative numbers of single, double, triple, and aromatic bound, number of ring, number of ring divided by the number of atoms or bonds, number of benzene ring, number of benzene ring divided by the number of atom, molecular weight and average molecular weight.

Burden eigenvalues descriptors defined as eigenvalues of a modified connectivity matrix, which could be called burden matrix B., the B matrix representing an H-depleted molecular graph. BCUT descriptor (Burden-CAS-University of Texas eigenvalues) are based on a significant extension of Burden approach, considering three classes of matrices whose diagonal elements correspond to 1) atomic charge-related values, 2) atomic polarizability-related values, and 3) atomic H-bond abilities.³³

Topological descriptors are based on a graph representation of the molecule. They are numerical quantifiers of molecular topology obtained by the application of algebraic operators to matrices representing molecular graphs and whose values are independent of vertex numbering or labeling. They can be sensitive to one or more structural features of the molecule such as size, shape, symmetry, branching and cyclicity and can also encode chemical information concerning atom type and bond multiplicity. Balaban index is a variant of connectivity index, represents extended connectivity and is a good descriptor for the shape of the molecules and modifying biological process. Nevertheless, some of chemists have used this index successfully in developing QSPR/QSRR models.

Charge descriptors were defined in terms of atomic charges and used to describe electronic aspects both of the whole molecule and of particular regions, such atoms, bonds, and molecular fragments. Charge descriptor calculated by computational chemistry and therefore can be consider among quantum chemical descriptor. Electrical charges in the molecule are the driving force of electrostatic interactions, and it is well known that local electron densities or charge play a fundamental role in many chemical reactions, physic-chemical properties and receptors-ligand binding affinity.³⁴

CONCLUSION

In this study, an accurate QSRR model for estimating the retention indices (RI) of essential oils from five Pimpinella species gathered from central Turkey which obtained by GC and GC-MS was developed by employing the linear model (GA-PLS) and nonlinear model (GA-KPLS). The most important molecular descriptors selected represent the electronic, charge and constitutional descriptors that are known to be important in the retention mechanism of essential oils. Two models have good predictive capacity and excellent statistical parameters. The results obtained by GA-KPLS were compared with the results obtained by GA-PLS model. The results demonstrated that GA-KPLS was more powerful in the retention indices of the essential oils than GA-PLS. This model could accurately predict the partitioning of these components that did not exist in the modeling procedure. It was easy to notice that there was a good prospect for the GA-KPLS application in the QSRR modeling.

Recebido em 10/1/11; aceito em 19/3/11; publicado na web em 10/6/11

• 1. Burt, S.; Food Microbiol 2004, 94, 223.
• 2. Kim, N. S.; Lee. D. D.; J. Chromatogr., A 2002, 982, 31.
• 3. Eminagaoglu, O.; Tepe, B.; Yumrutas, O.; Akpulat, H. A.; Daferera, D.; Polissiou, M.; Sokmen, A.; Food Chem 2007, 100, 339.
• 4. Mathew, V. A.; Pimpinella, L.; Flora of Turkey and East Aegean Islands, Edinburgh University Press: Edinburgh, 1972, vol. 4.
• 5. Baytop, T.; Turkiyede Bitkilerle Tedavi, Nobel Tip Kitapevi, Istanbul, 1999.
• 6. Joulain, D.; Konig W. A.; The Atlas of Spectra Data of Sesquiterpene Hydrocarbons, E. B. Verlag: Hamburg, 1998.
• 7. Jennings, W.; Shibamoto, T.; Quantitative Analysis of Flavor and Fragrance Volatile by Glass Capillary Column Gas Chromatography, Academic Press: New York, 1980.
• 8. Ong, V. S.; Hites, R. S.; Anal. Chem 1991, 63, 2829.
• 9. Peng, C. T.; Ding, S. F.; Hua, R. L.; Yang, W. C.; J. Chromatogr 1988, 436, 137.
• 10. Kaliszan, R.; Structure and Retention in Chromatography, Harwood: Amsterdam, 1997.
• 11. Qin, L. T.; Liu, Sh. Sh.; Liu, H. L.; Tong, J.; J. Chromatogr., A 2009, 1216, 5302.
• 12. Noorizadeh, H.; Farmany, A.; Noorizadeh, M.; Quim. Nova 2011, in press.
• 13. Noorizadeh, H.; Farmany, A.; Chromatographia 2010, 72, 563.
• 14. Sauer-Leal, E.; Okada, F. M.; Peralta-Zamora, P.; Quim. Nova 2008, 31, 1621.
• 15. Kaliszan, R.; J. Chromatogr., A 1993, 656, 417.
• 16. Lindgren, F.; Geladi, P.; Wold, S.; J. Chemometrics 1993, 7, 45.
• 17. Braak, J. M.; Birks, H. J. B.; J. Chemometrics 1994, 8, 169.
• 18. Krämer, N.; Boulesteix, A. L.; Tutz, G.; Chemom. Intell. Lab. Syst 2008, 94, 60.
• 19. Tabanca, N.; Demirci, B.; Ozek, T.; Kirimer, N.; Can Baser, K. H.; Bedir, E.; Khand, I. A.; Wedge, D. E.; J. Chromatogr., A 2006, 1117, 194.
• 20. Todeschini, R.; Consonni, V.; Mauri, A.; Pavan, M.; DRAGON-Software for the calculation of molecular descriptors, version 3.0 for Windows, 2003.
• 21. Costa Filho, C. A.; Poppi, R. J.; Quim. Nova 1999, 22, 405.
• 22. Lemes, M. R.; Pino Júnior, A. D.; Quim. Nova 2002, 25, 539.
• 23. Depczynski, U.; Frost, V. J.; Molt, K.; Anal. Chim. Acta 2000, 420, 217.
• 24. Samistraro, G.; Muniz, G. I. B.; Peralta-Zamora, P.; Cordeiro, G. A.; Quim. Nova 2009, 32, 1422.
• 25. NicolaÏ, B. M.; Theron, K. I.; Lammertyn, J.; Chemom. Intell. Lab. Syst 2007, 85, 243.
• 26. Rosipal, R.; Trejo, L. J.; J. Mach. Learning Res 2001, 2, 97.
• 27. Noorizadeh, H.; Farmany, A.; J. Chin. Chem. Soc. 2010, 57, 1.
• 28. Afantitis, A.; Melagraki, G.; Sarimveis, H.; Koutentis, P. A.; Markopoulos, J.; Igglessi-Markopoulou, O.; Bioorg. Med. Chem 2006, 14, 6686.
• 29. Golbraikh, A.; Tropsha, A.; J. Mol. Graphics Modell 2002, 20, 269.
• 30. Kim, K.; Lee, J. M.; Lee, I. B.; Chemom. Intell. Lab. Syst 2005, 79, 22.
• 31. Salvi, M.; Dazzi, D.; Pelistri, I.; Neri, F.; Wall, J. R.; Ophthalmology 2002, 109, 1703.
• 32. Kim, K.; Lee, J. M.; Lee, I. B.; Chemom. Intell. Lab. Syst 2005, 79, 22.
• 33. Booth, T. D.; Azzaoui, K.; Wainer, I. W.; Anal. Chem 1997, 69, 3879.
• 34. Todeschini, R.; Consonni, V.; Handbook of molecular descriptors, Wiley-VCH: Weinheim, 2000.

Publication Dates

• Publication in this collection
  06 Oct 2011
• Date of issue
  2011

History

• Received
  10 Jan 2011
• Accepted
  19 Mar 2011