The molecular retention mechanism in reversed-phase liquid chromatography of meso-ionic compounds by...

The Molecular Retention Mechanism in Reversed-Phase LiquidChromatography of Meso-ionic Compounds by Quantitative Structure ±

Retention Relationships (QSRR)

by Ta√nia M. G. Almeidaa), Andrei Leitaƒoa), Maria Luiza C. Montanarib), and Carlos A. Montanari*c)

a) Nu¬ cleo de Estudos em QuÌmica Medicinal ± NEQUIM ± Departamento de QuÌmica,Universidade Federal de Minas Gerais, Campus Pampulha, 31270-901, Belo Horizonte, MG, Brazil

b) Departamento de QuÌmica, Universidade Federal de Saƒo Carlos, Rodovia Washington Luiz, Km 235,13565-905, Saƒo Carlos, SP, Brazil

c) Grupo de Estudos emQuÌmica Medicinal de Produtos Naturais ± NEQUIMED-PN ± Instituto de QuÌmica deSaƒo Carlos, Universidade de Saƒo Paulo, Av. Trabalhador Sancarlense 400, 13560-970, Saƒo Carlos, SP, Brazil

(phone: � 55-16-3373-9986; fax: � 55-16-3373-9985; e-mail : [email protected])

Absorption, distribution, metabolism, and excretion (ADME) properties are of invaluable importance to abioactive compound. Permeation process is one of the most widely studied by many different techniques.Among them, reversed-phase liquid chromatography (RP-LC) is proving to be of great interest, in whichretention time can provide some insights on the hydrophobic interaction behavior. A series of antiprotozoalmeso-ionic 1,3,4-thiadiazolium-3-aminides were studied in order to obtain the retention-time data through twodifferent organic modifiers using a commercially available Supelcosil ABZ� Plus column. MeOH/buffer mobilephase provided the best results in assessing the hydrophobicity of these compounds as compared to MeCN.Quantitative Structure ±Retention Relationships (QSRR) models were built to identify the physicochemicalproperties that govern the process by using PLS analysis. VolSurf descriptors extracted from water (H2O),hydrophobic (DRY), amide N-atom (N(1)) and carbonyl O-atom (O) probes showed that the retention ispositively correlated to the molecules size and shape, hydrophobic pattern, and dispersion forces. Hydrogenbonding depicts the solvent force considering the donor and acceptor differences between MeOH and MeCNorganic modifiers.

Introduction. ± Cyclic polyheteroatom betaines are meso-ionic compounds withsubstantial charge separation and with a valuable array of biological activities [1]. Somebiological activities, ranging from antibacterial [2], antitumoral [3] to more recentlyantileishmaniasis [4], have been described for these compounds. Pharmacodynamicmechanisms of this wide spectrum of action are not known, but Shinzatto et al. [5]suggested that this charge mobility, in association with a small polyheteroatomicsystem, have a high probability of strong interactions with biomacromolecules such asDNA and proteins.

Though the pharmacodynamic process is important, the pharmacotherapeuticprofile is also dependent on pharmacokinetics (described by Absorption, Distribution,Metabolism, and Excretion (ADME)) [6] [7]. ADME Properties [8] are very difficultto be determined and yet very time-consuming, even when using in vitro models.

Permeability [9] is one of the key descriptors to be considered in a molecule whenlooking for drug-like properties [10]. Although permeability and absorption mecha-nisms are not completely known, their relations to pharmacological profiles are verywell described, especially when octan-1-ol/H2O partition coefficient (log Poct) is used as

CHEMISTRY & BIODIVERSITY ± Vol. 2 (2005) 1691

¹ 2005 Verlag Helvetica Chimica Acta AG, Z¸rich

a model [11 ± 13]. log Poct Values can be calculated by many in silico methods [14], butthey have low predictive power when zwitterionic molecules are considered.Chromatographic methods [15], such as reversed-phase liquid chromatography (RP-LC) [16], are able to describe this property in a fast way, with good reproducibility andwithout the need of sample purity. The log kw value is widely used to describe theretention behavior between mobile and stationary phases. Among many conventionalalkylsiloxane-bonded phases, the Supelcosil LC-ABZ column is one of the mostplausible candidates as a model for the octan-1-ol/H2O partition system [17].

Correlation of RP-LC retention times with compound properties can be achievedby Quantitative Structure ±Retention Relationship (QSRR) [18] applying eitherMultiple Linear Regression (MLR) or Partial Least Square (PLS) analyses. They arealso useful for the establishment of extra-thermodynamic relationships [19], by whichmolecular descriptors (such asVolSurf parameters) [20] can be useful for modelling thepartitioning process and other pharmacokinetic properties [21]. The main drawback ofthe PLS technique [22] is that these latent variables (LV) containing a huge number ofdescriptors are not prone to ease chemical interpretation. The VolSurf program can atleast partially solve this problem compressing the information present in the GRIDMolecular Interaction Fields (MIF) into few numerical descriptors that are simple tounderstand and to interpret. Also, the Fractional Factorial Design (FFD) procedure ishelpful to deal with this problem by reducing the number of descriptors that isconsidered noise [23].

This paper is aimed at: i) evaluating the lipophilic character of a previouslydescribed series of nine in-house synthesized meso-ionic compounds [24] (Fig. 1) byisocratic RP-LC analysis. The mobile phases consisted of buffer, and MeOH or MeCNas the organic modifiers eluting in the stationary phase Supelcosil LC-ABZ�Pluscolumn; ii) establishing QSRR to explain the dependence of chemical structure uponretention of the studied compounds, and also to clarify the role played by the twoorganic modifiers. Since there is no counterpart in the literature, this will allow us toevaluate the structure ± retention data to calculate partition coefficients for unsynthe-sized and untested new polyheteroatom betaines with antileishmaniasis activity.

CHEMISTRY & BIODIVERSITY ± Vol. 2 (2005)1692

Fig. 1. Structure of the meso-ionic 1,3,4-thiadiazolio-2-aminides

a) Lipophilic substituent constant � fromoctan-1-ol system (anilines). b) Lipophilicsubstituent constant � from RP-LC MeOH/buffer (meso-ionic).

Results and Discussion. ± Meso-ionic compounds were chromatographed withdifferent organic modifier concentrations (25 :75 (v/v)), and the logarithms of theretention factors on the ABZ� Plus column were extrapolated to 100% aqueous phase(log kw), using MeOH or MeCN (Table).

The Table shows that the slopes (S) related to the organic modifier concentration(Eqn. 2, in the Exper. Part) are negative in all cases. Regarding the strength of thesolvent, it can be observed that the plot of �S vs. log kw for the MeOH/buffer systemhas a good linear fit and high prediction power according to compounds similar in H-bonding pattern (r2� 0.924, Q2

LOO� 0.879). This trend is not so clear when MeCN/buffer is the mobile phase, and a quite poor relationship is established (r2� 0.570,Q2LOO� 0.414). When the mobile phase is MeOH/buffer, compounds 3 and 8 (Fig. 1)

are less retained, whereas all others with lipophilic groups, i.e., Me, Ph, t-Bu, Pr (2, 6, 7,and 9, resp.) are more retained than the parent compound 1 (Table). These propertiesrelated to the partition process are characterized and discussed with more details in thesection QSRR Studies.

The lipophilic substituent constant (�) from [11] was used to identify thehydrophobic contribution to partitioning of analytes into stationary phase fromMeOH/buffer and octan-1-ol/H2O coefficient from log P determination. Since there isno previous counterpart for such comparison, the octan-1-ol/H2O lipophilic constantsobtained for p-anilines were used as standard. According to MLR analysis there is agood correlation between �MeOH (RP-LC-ABZ) and �oct (Fig. 2 and Eqn. 1), showingthat there is a similar chemical environment between these two systems.

�MeOH� 0.747(�0.141), �oct� 0.085(�0.163) (1)

n� 9, r2� 0.888, s� 0.287; F� 55.655; Q2LOO� 0.791, Q2

LTO� 0.729

QSRR Studies. VolSurf descriptors were generated from the data set based on fourchemical probes: water (H2O), hydrophobic (DRY), carbonyl O-atom (O) and amide

Table. Logarithms of the Retention Factors on ABZ� Plus Columns Extrapolated to 100% Aqueous Phase(log kw)

Compound MeOH/buffera) MeCN/buffera)

log kw S Q2 log kw S Q2

1 3.203 � 0.0437 0.846 3.149 � 0.0414 0.8752 3.301 � 0.0434 0.898 3.339 � 0.0426 0.9053 2.935 � 0.0338 0.935 3.071 � 0.0418 0.8224 3.618 � 0.0435 0.939 3.636 � 0.0465 0.8525 3.644 � 0.0430 0.934 3.601 � 0.0454 0.8506 4.973 � 0.0569 0.942 4.124 � 0.0493 0.8817 4.441 � 0.0536 1.000 3.618 � 0.0490 0.8608 2.705 � 0.0372 0.951 3.251 � 0.0442 0.9249 4.698 � 0.0552 0.818 3.657 � 0.0417 0.934

a) S: Slope of organic modifier concentration (�). Q2 : Cross-validation correlation coefficient.


N-atom (N1). PLS Analyses were carried out, and the following models weregenerated after refinement by the FFD procedure:

MeOH/buffer (Fig. 3,a):

n� 9, r2� 0.978, SDEC� 0.113, Q2LOO� 0.962 (Q2

LTO� 0.947), SDEPLOO� 0.147(SDEPLTO� 0.174);

MeCN/buffer (Fig. 3,b):

n� 9, r2� 0.887, SDEC� 0.104, Q2LOO� 0.834 (Q2

LTO� 0.817), SDEPLOO� 0.125(SDEPLTO� 0.132)

Fig. 3,a represents the relationship between MeOH/buffer log kw (Y scores) andcalculated VolSurf descriptors (X scores) from the set of meso-ionic compounds,analyzed by PLS projection to Latent Variable 1 (LV1). Compounds with hydrophobicR groups (2, 6, 7, and 9) had a longer retention time than the more hydrophilic ones (8

Fig. 3. PLS Plot of the compound set. a) MeOH, b) MeCN. Y Scores� log kw; X scores�VolSurf descriptors .


Fig. 2. Linear relationship between �MeOH and �oct formeso-ionic compounds

and 3). MeCN (Fig. 3,b) did not show the same hydrophobic trend. Hydrophobic andhydrophilic properties are clearly described by the LV1 that was used to build theMeOH/buffer PLS model (Fig. 3). This LV1 is composed by ten descriptors (see [20]for an in-depth description), depicted in Fig. 4,a.

Partition process is basically governed by bulkiness and shape, H-bond, electro-statics, and hydrophobic interactions, and the following terms are grouped according tothese properties.

Fig. 4. PLS Analyses of variables from MeOH-buffer system. a) PLS Weights, b) meso-ionic 6 profile, c) meso-ionic 8 profile. Probes: H2O (water); DRY (hydrophobic); O (carbonyl O-atom); N1 (amide N-atom).Variables: molecular surface (S); globularity (G); polarizability and dispersion forces (W1); −integy× moment

(Iw); capacity factor (Cw); hydrophobic region (D); H-bonding (HB).


Molecular surface (S) and globularity (G) are positively related to the retention(Fig. 4,a). They are related to the molecule size and shape, respectively. Compoundswith no substituent (like 1) or small ones (compound 8 ; Fig. 4,c) have lower values ofglobularity than those with bulkier branches (compounds 6 (Fig. 4,b), 7, and 9). Theretention profile described byG is in good agreement with the cavitation theory, wheresphere molecules are less retained because they are not as flexible as the flat ones.Molecular surface is related to G because higher branched molecules have a biggersurface.

The hydrophobic core is expressed by a descriptor with the highest weight(D1(DRY); Fig, 4,a). Molecules with hydrophobic side chains (like 6) are moreretained than hydrophilic ones (like 8), as it can be seen by MIF analysis for molecules6 and 8 (Fig. 5,a and b, resp.).

Polarizability and dispersion forces (W1(O), W1(N1)) related to weaker inter-actions are important to the permeation process, and they are positively correlated tothe retention times (Fig. 4,a). The energy level comprised by the W1 region is at� 0.2 kcal mol�1.

The size of hydrophilic regions at a higher cutoff (�2.0 kcal mol�1) is described bythe capacity factor (Cw4(H2O)). This variable is detrimental to the compound×s

Fig. 5. Representation of molecular interaction fields for compounds 6 and 8. Hydrophobic fields for compounda) 6 and b) 8. Hydrophilic fields for c) 6 and d) 8. Hydrophilic integy moments are shown as vectors in c and d.

MIF Cut-off: hydrophobic (�1 kcal mol�1); hydrophilic (� 3 kcal mol�1).


retention (Fig. 4,a). Therefore, molecules with large hydrophilic regions are lessretained than the others.

−Integy× moments from H2O probe (Iw5(H2O) and Iw6(H2O) at � 3.0 and� 4.0 kcal mol�1 energy levels, resp.) are unfavourable to the compound×s retention.Vectors pointing from the barycentre to hydrophilic regions of a molecule representthese descriptors. Fig. 5,c shows that molecule 6, the most retained one, has smallvectors because its hydrophilic atoms are nearby the barycentre. The less retained one,compound 8, has high −integy× moments pointing to the OH group in the para-position,very far from its centre of mass (Fig. 5,d).

Hydrogen-bonding variables (HB1(N1) and HB2(N1)) are detrimental to theretention (Fig. 4,a) because these interactions broadly occur between the solute andmobile phase. Only the N1 probe is responsible for the H-bonding in this model. Thisprobe is essentially H-bond donor, which simulates the H-bonding interaction betweenMeOH and two weak H-bond acceptors (MeO and OH substituents) [25]. That profileis far different from MeCN PLS weights (Fig. 6), because this organic modifier doesnot have an acidic H-atom in its structure. This information is helpful to understandwhy compounds 8 and 3 did not have a large difference among the retention valueswhen MeCN was used [26].

Conclusions. ± Statistically significant 3D-QSRR models were developed based onRP-LC. The results obtained on a Supelcosil LC-ABZ� Plus column provided suitablemodels for estimating log kw values with MeOH/buffer mobile phase. However, MeCNhas lower sensitivity to changes in the compound structures, and it is not appropriate asan organic modifier in the study of the retention behavior of these molecules.

VolSurf analysis provided an easy way to integrate 2D and 3D information, becausei) it was straightforward to analyse a huge amount of 3D data by a small number of 2Dparameters, and ii) 2D descriptors were better defined by 3DMIF in a backward mode.


Fig. 6. PLS Variable analysis from MeCN/buffersystem. Probes: H2O (water); DRY (hydropho-bic); O (carbonyl O-atom); N1 (amide N-atom). Variables: hydrophilic regions (W);hydrophobic region (D); amphiphilic moment

(A); molecular weight (MW).

Statistical overfitting was avoided by using only the first latent variable (LV1) fromPLS analyses. Although we have used LOO and LTO procedures to internally validatethe data, we also divided them into a training and a test set. Thorough training wascarried out, and we obtainedQ2� 0.934 for the best and 0.646 for the worst set (leaving33% for each test set trial). Residuals were in the range of 0.11 and 0.22, respectively.These show that predictions are not far beyond the range of the training set, and thuscircumvent the data overfitting. The partition coefficient was described adequatelyfromMeOH/buffer system by a QSRR study on this meso-ionic dataset. Size and shapedescriptors together with hydrophobic and electrostatic interactions are governing theretention mechanism. The role of H-bonding is also raised for explanation comprisingthe differences between the two organic modifiers.

In the data set employed, it should be noted that the bulk of the structural diversitylies in the nature of the substituent linked to the 4-position of the aniline-like ring.Nevertheless, the compounds studied display the chromatographic partitioning in abroad range of 2.3 log units (acceptably distributed across the range of values), whichapplies for the chosen MeOH/buffer system. However, the compounds studied weresynthesized by the appropriate choice of lipophilic constants, �oct , a more nearlysuitable data set for statistical modelling would require more structural diversity and alarger number of compounds. However, the correct selection of this small number ofcompounds to establish the relationships depicted in this paper is of utmost importanceto serve as pre-filter in the search for new compounds to be synthesized. Overall, it canbe anticipated that many chromatographic partition coefficients could be automated inorder to allow an array of −well-measured× and reliable pharmacokinetic data thatshould be of support for early access of pharmacokinetic properties in the drug-designarena.

Experimental Part

Instruments. The experiments were performed on a Shimadzu instrument equipped with two pumps (modelLC-10AD) with a variable-wavelength detector (model SPD-6AV) operating in a range 220 ± 300 nm coupledwith an integrator (model LC-R6A). The injecting valve was fitted with a 20-�l sample loop. The commerciallyavailable Supelcosil LC-ABZ� Plus column was used in all experiments.

Materials. MeOH and MeCN were purchased from Carlo Erba (HPLC-grade). Morpholine-4-propane-sulfonic acid (MOPS), NaNO3, and NaOH were obtained from Merck. Milli-Q water (Milli-Q UF Plus system,Millipore Corporation) was used to prepare mobile phases that were degassed by ultrasonication immediatelybefore use. The meso-ionic compounds 1 ± 9 were synthesized, purified, and identified according to theprocedure of Britto [24], and their structures are given in Fig. 1.

RPLC Procedure. The meso-ionic compounds were dissolved in MeOH, and the injection volume was 20 �lfor all experiments. Chromatographic runs were carried out at r.t. (27� 2�); the flow-rate was 1 ml min�1. Themobile phase consisted of 25 m�MOPS buffer (�� 30 m�), adjusted to pH 7.4 by the addition of 1.0� NaOH,and MeOH or MeCN as org. modifier. Chromatographic retention data are expressed by the logarithm ofretention factor, log k, defined as log k� log [(tr� t0)/t0], where tr and t0 are the retention times of the compoundand a non-retained compound (uracyl), resp. All retention factors given represent the mean of 2 ± 4determinations of each sample. There was a good reproducibility of the retention times, varying 1.2% (RSD)during the period of experiments. These compounds were eluted with mobile phases containing MOPS bufferand org. modifier fractions (�) ranging from 25 to 75% (v/v). Linear relationships between log k and � valueswere found for these compounds. log kw, extrapolated when � organic modifier is set to zero, was obtained as thelipophilic index. The values of log kw and S were taken by linear regression using:

log k� log kw�S � (2)


Structure Optimization and VolSurf Method. 3D Structures of the meso-ionic compounds were drawnstarting from X-ray crystallographic data for compound 1 taken as template from CCDC (CSD code:JOSWUJ). The molecular optimization procedure was carried out in Sybyl 6.8 package, on a Silicon GraphicsOctane R12000 workstation.Gasteiger ± H¸ckel charges were calculated before the geometry optimization usingTripos force field and 1000 Steepest-Descend (SD) steps.

Molecular-interaction fields (MIF) were calculated using hydrophilic (H2O), hydrophobic (DRY), H-bond acceptor (carbonyl O-atom (O)) and H-bond donor (amide N-atom (N1)) probes that are available in theGRID program, and converted to VolSurf descriptors automatically on a PC Linux 1.2 GHz.

Statistical Methods. The multilinear regression analysis (MRL) was performed on partition parametersfrom RP-LC, using the TSAR program (version 3.3). In PLS regression, the X block of independent variables(descriptors) is correlated with the Y vector (activity) in such a way that the projected coordinates T are goodpredictors of Y. In this case, the retention factors are included in the decomposition procedure. The goal of PLSregression is to build a model that will be able to predict Y from X, and to describe their common structure. Thebest model obtained should be useful to calculate reliable predictions of Y values for new molecules, notincluded in the model. It is important to note that the PLS regression method can deal with the kind X matricesused in 3D-QSRR studies. In the present case, the X matrix contains much less objects (molecules) thanvariables, and the use of multiple linear regression (MLR) is not appropriate. PLS decomposes the X matrix asthe product of two smaller matrices, P and T, which leads to the latent variables (LVs) that represent thestructure of the X and Ymatrices, maximizing the fitting between the Xs and Ys. Selecting the correct number ofLVs (dimensionality) and testing the predictive power of the model is of critical importance in PLS. PLSAnalyses were from VolSurf program, and a pre-treatment procedure was applied on these sets of auto-scaledparameters from MeCN and MeOH as organic modifiers by the application of the Fractional Factorial Design(FFD). The final models for MeOH and MeCN (ten variables each one) were generated by PLS weights andregression parameters analysis. The first Latent Variable (LV1) was used to build the models, and the predictivepower was determined using a Leave-One-Out (LOO) and Leave-Two-Out (LTO) cross-validation procedures.In addition, the substituted template scaffold was divided into training and test sets. One third of all compounds(three molecules) were held out for testing. This kind of external validation is also appropriate to tackle thestatistical overfitting, and minimizes it by showing that the generated model can predict the never seenchromatographic retention factor.

REFERENCES

[1] C. A. Montanari, J. P. B. Sandall, Y. Miyata, J. Miller, J. Chem. Soc., Perkin Trans. 2 1994, 2571; A. C. S. dosSantos, A. Echevarria, Magn. Reson. Chem. 2001, 39, 182.

[2] C. A. Montanari, A. T. Amaral, A. Giesbrecht, Pharm. Sci. 1997, 3, 565.[3] N. Grynberg, A. C. Santos, A. Echevarria, Anti-Cancer Drug Des. 1997, 8, 88.[4] E. F. Silva, M. M. Canto-Cavalheiro, V. R. Braz, L. Cysnei-Finkelstein, L. L. Leon, A. Echevarria, Eur. J.

Med. Chem. 2002, 37, 979.[5] T. O. Shinzato, N. F. Grynberg, R. M. Gomes, A. Echevarria, J. Miller, Med. Sci. Res. 1989, 17, 856.[6] a) H. van deWaterbeemd, D. A. Smith, K. Beaumont, D. K.Walker, J. Med. Chem. 2001, 44, 1313; b) C. A.

Lipinski, J. Pharm. Toxicol. Methods 2000, 44, 235; c) C. A. Lipinski, F. Lombardo, B. W. Dominy, P. J.Feeney, Adv. Drug Deliv. Rev. 1997, 23, 3.

[7] H. Yu, A. Adedoyin, Drug Discov. Today 2003, 8, 852.[8] a) A. M. Davis, R. J. Riley, Curr. Opin. Chem. Biol. 2004, 8, 378; b) G. Klopman, L. R. Stefan, R. D.

Saiakhov, Eur. J. Pharm. Sci. 2002, 17, 253; c) S. Ekins, C. L. Waller, P. W. Swaan, G. Cruciani, S. A.Wrighton, J. H. Wikel, J. Pharm. Toxicol. Methods 2000, 44, 251.

[9] A. M‰lki‰, L. Murtom‰ki, A. Urtti, K. Kontturi, Eur. J. Pharm. Sci. 2004, 23, 13.[10] M. Br¸stle, B. Beck, T. Schindler, W. King, T. Mitchell, T. Clark, J. Med. Chem. 2002, 45, 3345.[11] T. Fujita, J. Iwasa, C. Hansch, J. Am. Chem. Soc. 1964, 86, 5175.[12] P. Buchwald, N. Bodor, Curr. Med. Chem. 1998, 5, 353.[13] H. van de Waterbeemd, B. Testa, in −The Parametrization of Lipophilicity and the Structural Properties in

Drug Design×, Ed. B. Testa, Academic Press, London, 1987, Vol. 16, p. 85.[14] R. Mannhold, G. Cruciani, K. Dross, R. Rekker, J. Comput.-Aided Mol. Des. 1998, 12, 573.[15] S. K. Poole, C. F. Poole, J. Chromatogr., B 2003, 797, 3; M. Amato, F. Barbato, P. Morrica, F. Quaglia,

M. I. L. Rotonda, Helv. Chim. Acta 2000, 83, 2836; K. Valko¬ , J. Chromatogr., A 2004, 1037, 299.


[16] M. M. Britto, Q. B. Cass, I. C. S. F. Jardim, C. A. Montanari, H. Y. Aboul-Enein, Chromatographia 2001,53, 11; A. Vailaya, C. Horva¬th, J. Chromatogr., A 1998, 829, 1.

[17] N. C. Dias, M. I. Navas, C. F. Poole, Analyst 2003, 128, 427.[18] R. Kaliszan, J. Chromatogr. A 1993, 656, 417; M. P. Montanƒ a, N. B. Pappano, N. B. Debattista, J. Raba, J. M.

Luco, Chromatographia 2000, 51, 727; M. Jezierska, I. Cendrowska, M. Markuszewski, R. Kaliszan, B.Buszewski, Chromatographia 2000, 51, 111; J. M. Luco, A. P. Salinas, A. A. J. Torrieiro, R. N. Va¬zquez, J.Raba, E. Marchevsky, J. Chem. Inf. Comput. Sci. 2003, 43, 2129; L. I. Nord, D. Fransson, S. P. Jacobsson,Chemom. Intell. Lab. Syst. 1998, 44, 257.

[19] M. L. C. Montanari, C. A. Montanari, D. P. Veloso, A. E. Beezer, J. C. Mitchell, P. L. O. Volpe, Quant.Struct.-Act. Relat. 1998, 17, 102.

[20] G. Cruciani, M. Pastor, W. Guba, Eur. J. Pharm. Sci. 2000, 11, Suppl. 2, S29; G. Cruciani, P. Crivori, P. A.Carrupt, B. Testa, J. Mol. Struct.-THEOCHEM 2000, 503, 17.

[21] M. L. C. Montanari, A. D. Andricopulo, C. A. Montanari, Thermochim. Acta 2004, 417, 283; I. Zamora, T.Oprea, G. Cruciani, M. Pastor, A.-L. Ungell, J. Med. Chem. 2003, 46, 25.

[22] R. Put, C. Perrin, F. Questier, D. Coomans, D. L. Massart, Y. Vander Heyden, J. Chromatogr., A 2003, 988,261.

[23] VolSurf 3.0.11 manual. Molecular Discovery, U.K.[24] M. M. Britto, Ph.D. Thesis, Federal University of Minas Gerais, Brazil, 1999.[25] I. Nobeli, S. L. Yeoh, S. L. Price, R. Taylor, Chem. Phys. Lett. 1997, 280, 196; R. Parthasarathi, V.

Subramanian, N. Sathyamurthy, J. Phys. Chem. A 2005, 109, 843; J. Reynisson, E. McDonald, J. Comput.-Aided Mol. Des. 2004, 18, 421; D.-S. Ahn, S.-W. Park, S. Lee, B. Kim, J. Phys. Chem. A 2003, 107, 131.

[26] E. S. Kryachko, M. T. Nguyen, J. Phys. Chem. A 2002, 106, 4267.

Received July 27, 2005


The molecular retention mechanism in reversed-phase liquid chromatography of meso-ionic compounds by...

Documents

Transcript of The molecular retention mechanism in reversed-phase liquid chromatography of meso-ionic compounds by...