Quantitative structure-enantioselective retention relationships for chromatographic separation of arylalkylcarbinols on Pirkle type chiral stationary phases

Intelligent Recommendation Systems Powered by Consensus Neural Networks: The Ultimate Solution for Finding Suitable Chiral Chromatographic Systems?

The selection of suitable combinations of chiral stationary phases (CSPs) and mobile phases (MPs) for the enantioresolution of chiral compounds is a complex issue that often requires considerable experimental effort and can lead to significant waste. Linking the structure of a chiral compound to a CSP/MP system suitable for its enantioseparation can be an effective solution to this problem. In this study, we evaluate algorithmic tools for this purpose. Our proposed consensus model, which uses multiple optimized artificial neural networks (ANNs), shows potential as an intelligent recommendation system (IRS) for ranking chromatographic systems suitable for the enantioresolution of chiral compounds with different molecular structures. To evaluate the IRS potential in a proof-of-concept stage, 56 structural descriptors for 56 structurally unrelated chiral compounds across 14 different families are considered. Chromatographic systems under study comprise 7 cellulose and amylose derivative CSPs and acetonitrile or methanol aqueous MPs (14 chromatographic systems in all). The ANNs are optimized using a fit-for-purpose version of the chaotic neural network algorithm with competitive learning (CCLNNA), a novel approach not previously applied in the chemical domain. CCLNNA is adapted to define the inner ANN complexity and perform feature selection of the structural descriptors. A customized target function evaluates the correctness of recommending the appropriate CSP/MP system. The ANN-consensus model exhibits no advisory failures and requires only an experimental attempt to verify the IRS recommendation for complete enantioresolution. This outstanding performance highlights its potential to effectively resolve this problem.

The Composition of the Mobile Phase Affects the Dynamic Chiral Recognition of Drug Molecules by the Chiral Stationary Phase

More than half of all pharmaceuticals are chiral compounds. Although the enantiomers of chiral compounds have the same chemical structure, they can exhibit marked differences in physiological activity; therefore, it is important to remove the undesirable enantiomer. Chromatographic separation of chiral enantiomers is one of the best available methods to get enantio-pure substances, but the optimization of the experimental conditions can be very time-consuming. One of the most widely used chiral stationary phases, amylose tris(3,5-dimethylphenyl carbamate) (ADMPC), has been extensively investigated using both experimental and computational methods; however, the dynamic nature of the interaction between enantiomers and ADMPC, as well as the solvent effects on the ADMPC-enantiomer interaction, are currently absent from models of the chiral recognition mechanism. Here we use QM/MM and molecular dynamics (MD) simulations to model the enantiomers of flavanone on ADMPC in either methanol or heptane/2-propanol (IPA) (90/10) to elucidate the chiral recognition mechanism from a new dynamic perspective. In atomistic MD simulations, the 12-mer model of ADMPC is found to hold the 4/3 left-handed helical structure in both methanol and heptane/IPA (90/10); however, the ADMPC polymer is found to have a more extended average structure in heptane/IPA (90/10) than in methanol. This results from the differences in the distribution of solvent molecules close to the backbone of ADMPC leads to changes in the distribution of the (φ, ψ) dihedral angles of the glycoside bond (between adjacent monomers) that define the structure of the polymer. Our simulations have shown that the lifetime of hydrogen bonds formed between ADMPC and flavanone enantiomers in the MD simulations are able to reproduce the elution order observed in experiments for both the methanol and the heptane/IPA solvent systems. Furthermore, the ratios of hydrogen-bonding-lifetime-related properties also capture the solvent effects, in that heptane/IPA (90/10) is found to make the separation between the two enantiomers of flavanone less effective than methanol, which agrees with the experimental separation factors of 0.9 versus 0.4 for R/S, respectively.

Mining Chromatographic Enantioseparation Data Using Matched Molecular Pair Analysis

We apply matched molecular pair (MMP) analysis to data from ChirBase, which contains literature reports of chromatographic enantioseparations. For the 19 chiral stationary phases we examined, we were able to identify 289 sets of pairs where there is a statistically significant and consistent difference in enantioseparation due to a small chemical change. In many cases these changes highlight enantioselectivity differences between pairs or small families of closely related molecules that have for many years been used to probe the mechanisms of chromatographic chiral recognition; for example, the comparison of N-H vs. N-Me analytes to determine the criticality of an N-H hydrogen bond in chiral molecular recognition. In other cases, statistically significant MMPs surfaced by the analysis are less familiar or somewhat puzzling, sparking a need to generate and test hypotheses to more fully understand. Consequently, mining of appropriate datasets using MMP analysis provides an important new approach for studying and understanding the process of chromatographic enantioseparation.

Spontaneous chiral symmetry breaking in early molecular networks

BACKGROUND

An important facet of early biological evolution is the selection of chiral enantiomers for molecules such as amino acids and sugars. The origin of this symmetry breaking is a long-standing question in molecular evolution. Previous models addressing this question include particular kinetic properties such as autocatalysis or negative cross catalysis.

RESULTS

We propose here a more general kinetic formalism for early enantioselection, based on our previously described Graded Autocatalysis Replication Domain (GARD) model for prebiotic evolution in molecular assemblies. This model is adapted here to the case of chiral molecules by applying symmetry constraints to mutual molecular recognition within the assembly. The ensuing dynamics shows spontaneous chiral symmetry breaking, with transitions towards stationary compositional states (composomes) enriched with one of the two enantiomers for some of the constituent molecule types. Furthermore, one or the other of the two antipodal compositional states of the assembly also shows time-dependent selection.

CONCLUSION

It follows that chiral selection may be an emergent consequence of early catalytic molecular networks rather than a prerequisite for the initiation of primeval life processes. Elaborations of this model could help explain the prevalent chiral homogeneity in present-day living cells.

Chiral recognition by enantioselective liquid chromatography: mechanisms and modern chiral stationary phases

An overview of the state-of-the-art in LC enantiomer separation is presented. This tutorial review is mainly focused on mechanisms of chiral recognition and enantiomer distinction of popular chiral selectors and corresponding chiral stationary phases including discussions of thermodynamics, additivity principle of binding increments, site-selective thermodynamics, extrathermodynamic approaches, methods employed for the investigation of dominating intermolecular interactions and complex structures such as spectroscopic methods (IR, NMR), X-ray diffraction and computational methods. Modern chiral stationary phases are discussed with particular focus on those that are commercially available and broadly used. It is attempted to provide the reader with vivid images of molecular recognition mechanisms of selected chiral selector-selectand pairs on basis of solid-state X-ray crystal structures and simulated computer models, respectively. Such snapshot images illustrated in this communication unfortunately cannot account for the molecular dynamics of the real world, but are supposed to be helpful for the understanding. The exploding number of papers about applications of various chiral stationary phases in numerous fields of enantiomer separations is not covered systematically.

Simple method for the prediction of the separation of racemates with high-performance liquid chromatography on Whelk-O1 chiral stationary phase

A simple method for the prediction of whether or not a racemate can be separated on a Whelk-O1 chiral stationary phase has been developed. In this approach, molecules are represented by counting the number of atom types of the neighbors spheres of the chiral center. A decision tree is then used to decide based on a few of these atom count descriptors whether a given racemate can be separated. High values of correct prediction were obtained, namely with more than 94% for training sets and of about 90% for cross-validation results. The same rate of correct prediction was also obtained on an external data set. The descriptors can be rapidly and easily retrieved by just counting the atom types around the chiral center by inspecting the chemical diagram of the molecule. Furthermore, the decision tree model can be applied through the use of a small set of rules that eventually predicts whether or not a racemate is separated. Due to its computational simplicity, the procedure is of interest for experimentalists that need to make rapid assessment of the separation without having to program or input complex formulas.

Investigation of retention behaviour of non-steroidal anti-inflammatory drugs in high-performance liquid chromatography by using quantitative structure–retention relationships

In this paper, a quantitative structure-retention relationship (QSRR) method is employed to model the retention behaviour in reversed-phase high-performance liquid chromatography of arylpropionic acid derivatives, largely used non-steroidal anti-inflammatory drugs (NSAIDs). Computed molecular descriptors and the organic modifier content in the mobile phase are associated into a comprehensive model to describe the effect of both solute structure and eluent composition on the isocratic retention of these drugs in water-acetonitrile mobile phases. Multilinear regression (MLR) combined with genetic algorithm (GA) variable selection is used to extract from a large set of computed 3D descriptors an optimal subset. Based on GA-MLR analysis, a five-dimensional QSRR model is identified. All the four selected molecular descriptors belong to the category of GEometry, Topology, and Atom-Weights AssemblY (GETAWAY) descriptors. The related multilinear model exhibits a quite good fitting and predictive performance. This model is further improved using an artificial neural network (ANN) learned by error back-propagation. Finally, the ANN-based model displays a remarkably better performance as compared with the MLR counterpart and, based on external validation, is able to predict with good accuracy the behaviour of unknown arylpropionic NSAIDs in the range of mobile phase composition of analytical interest (between 35 and 75% acetonitrile (v/v)).

Modelling of retention of pesticides in reversed-phase high-performance liquid chromatography: Quantitative structure-retention relationships based on solute quantum-chemical descriptors and experimental (solvatochromic and spin-probe) mobile phase descriptors

A quantitative structure-retention relationship (QSRR) analysis based on multilinear regression (MLR) and artificial neural networks (ANNs) is carried out to model the combined effect of solute structure and eluent composition on the retention behaviour of pesticides in isocratic reversed-phase high-performance liquid chromatography (RP-HPLC). The octanol-water partition coefficient and four quantum chemical descriptors (the total dipole moment, the mean polarizability, the anisotropy of the polarizability and a descriptor of hydrogen-bonding based on the atomic charges on acidic and basic chemical functionalities) are considered as solute descriptors. In order to identify suitable mobile phase descriptors, encoding composition-dependent properties of both methanol- and acetonitrile-containing mobile phases, the Kamlet-Taft solvatochromic parameters (polarity-dipolarity, hydrogen-bond acidity and hydrogen-bond basicity, pi*, alpha and beta, respectively) and the 14N hyperfine-splitting constant (aN) of a spin-probe dissolved in the eluent are examined. A satisfactory description of mobile phase properties influencing the solute retention is provided by aN and beta or alternatively pi* and beta. The two seven-parameter models resulting from combination of aN and beta, or pi* and beta, with the solute descriptors were tested on a set of 26 pesticides representative of 10 different chemical classes in a wide range of mobile phase composition (30-60% (v/v) water-methanol and 30-70% (v/v) water-acetonitrile). Within the explored experimental range, the acidity of the eluent, as quantified by alpha, is almost constant, and this parameter is in fact irrelevant. The results reveal that aN and pi*, that can be considered as interchangeable mobile phase descriptors, are the most influent variables in the respective models. The predictive ability of the proposed models, as tested on an external data set, is quite good (Q2 close to 0.94) when a MLR approach is used, but the modelling capability can be further improved using an artificial neural network.

Quantitative structure-(chromatographic) retention relationships

Since the pioneering works of Kaliszan (R. Kaliszan, Quantitative Structure-Chromatographic Retention Relationships, Wiley, New York, 1987; and R. Kaliszan, Structure and Retention in Chromatography. A Chemometric Approach, Harwood Academic, Amsterdam, 1997) no comprehensive summary is available in the field. Present review covers the period of 1996-August 2006. The sources are grouped according to the special properties of kinds of chromatography: Quantitative structure-retention relationship in gas chromatography, in planar chromatography, in column liquid chromatography, in micellar liquid chromatography, affinity chromatography and quantitative structure enantioselective retention relationships. General tendencies, misleading practice and conclusions, validation of the models, suggestions for future works are summarized for each sub-field. Some straightforward applications are emphasized but standard ones. The sources and the model compounds, descriptors, predicted retention data, modeling methods and indicators of their performance, validation of models, and stationary phases are collected in the tables. Some important conclusions are: Not all physicochemical descriptors correlate with the retention data strongly; the heat of formation is not related to the chromatographic retention. It is not appropriate to give the errors of Kovats indices in percentages. The apparently low values (1-3%) can disorient the reviewers and readers. Contemporary mean interlaboratory reproducibility of Kovats indices are about 5-10 i.u. for standard non polar phases and 10-25 i.u. for standard polar phases. The predictive performance of QSRR models deteriorates as the polarity of GC stationary phase increases. The correlation coefficient alone is not a particularly good indicator for the model performance. Residuals are more useful than plots of measured and calculated values. There is no need to give the retention data in a form of an equation if the numbers of compounds are small. The domain of model applicability of models should be given in all cases.

Theoretical analysis on retention behavior of pigments in reversed-phase high-performance liquid chromatographic (HPLC)

Quantitative structure-retention relationship (QSRR) models have been used successfully to predict and explain retention behavior of pigments in reversed-phase high-performance liquid chromatography (HPLC). The semi-empirical quantum chemical method (PM3) in Gaussian98 was employed to calculate a set of molecular descriptors of pigments. Using multiple linear regression (MLR), we obtained empirical functions with high correlation coefficient between retention times and quantum-chemical descriptors. This analysis indicated that the proposed QSRR models were satisfactory.

Quantitative structure-retention relationships of pesticides in reversed-phase high-performance liquid chromatography

In this paper, a quantitative structure-retention relationships (QSRR) method is employed to predict the retention behaviour of pesticides in reversed-phase high-performance liquid chromatography (HPLC). A six-parameter nonlinear model is developed by means of a feed-forward artificial neural network (ANN) with back-propagation learning rule. Accurate description of the retention factors of 26 compounds including commonly used insecticides, herbicides and fungicides and some metabolites is successfully achieved. In addition to the acetonitrile content, included to describe composition of the water-acetonitrile mobile phase, the octanol-water partition coefficient (from literature) and four quantum chemical descriptors are considered to account for the effect of solute structure on the retention. These are: the total dipole moment, the mean polarizability, the anisotropy of polarizability and a descriptor of hydrogen bonding ability based on the atomic charges on hydrogen bond donor and acceptor chemical functionalities. The proposed nonlinear QSRR model exhibits a high degree of correlation between observed and computed retention factors and a good predictive performance in wide range of mobile phase composition (40-65%, v/v acetonitrile) that supports its application for the prediction of the chromatographic behaviour of unknown pesticides. A multilinear regression model based on the same six descriptors shows a significantly worse predictive capability.

Solvation of the Whelk-O1 chiral stationary phase: A molecular dynamics study

Density functional theory calculations and molecular dynamics simulations are employed to explore the solvation of the Whelk-O1 chiral stationary phase. First, a semi-flexible representation of the Whelk-O1 selective molecule is extracted from an extensive series of B3LYP/6-311+ G(2d,p) calculations. The resulting model is used to build a chiral surface, including end-caps, for molecular dynamics study of the interface between solvent and Whelk-O1. Three solvent environments in common use for Whelk-O1 HPLC have been examined: a normal-phase solvent of n-hexane/2-propanol; a reversed-phase solvent of water/methanol; and a supercritical solvent of CO(2) and methanol. In each case, we analyze the interface with an emphasis on solvent composition and solvent hydrogen bonding to the Whelk-O1 selector.

Quinine carbamate chiral stationary phases: Systematic optimization of steric selector-selectand binding increments and enantioselectivity by quantitative structure-enantioselectivity relationship studies

A series of quinine carbamate-based chiral stationary phases (CSPs) differing solely in the carbamate residue have been devised and a congeneric set of N-3,5-dinitrobenzoyl (DNB) amino acids (AAs) was separated into enantiomers on these CSPs by HPLC using a buffered hydro-organic mobile phase. Some details on retention and chiral recognition mechanisms have been investigated by application of quantitative structure-property relationship (QSPR) studies using the linear free energy relationship methodology, i.e., the extrathermodynamic approach. Retention factors of the high affinity enantiomer (log k2) and enantioselectivities (log alpha) were correlated with Taft's steric parameter as structural descriptor for the variability in the carbamate and AA residues, and statistically significant QSPR models could be obtained. They confirmed that the variance in the dependent variable (log k2, log alpha) is mainly associated with the steric bulkiness of the selectors' carbamate and of the AA residues. The retention factor of the second eluted enantiomers and the enantioselectivity first increased with steric bulkiness of the carbamate and AA residues, which may be explained by stronger dispersive interactions. After reaching an optimum, the dependent variable, however, declined with further increase of the steric bulkiness of the substituents, probably because of steric hindrance. The variability of the retention factors of the first eluted enantiomer could not be explained by steric descriptors. Instead, it has become obvious that the retention arises mainly from interactions of the DNB-AA and the quinine carbamate backbone, as it turned out to be more or less constant and solely to a minute amount modulated by the carbamate residue and the AA side chain. The QSPR models were fully in agreement with an earlier postulated chromatographically and spectroscopically derived hypothetical selector-selectand binding model.

Enantiophore modeling in 3D-QSAR. A data mining application on Whelk-O1 chiral stationary phase

A combination of the enantiophore concept described in a previous study and a quantitative structure enantioselective relationship (QSER) based on partial least squares (PLS) analysis is presented. In the present study, a comprehensive approach for describing the enantioselective binding properties of the Whelk-O1 chiral HPLC receptor is achieved using molecular descriptors calculated by the GRID program. The GRID descriptors allow us to describe the molecules in terms of their ability to form favorable interactions with independent chemical groups (probes) that can be related to receptor sites. For each molecule, we compute 120 enantiophore descriptors representing the energy contributions from all possible pairwise combinations of probes. The overall procedure was simplified by considering only the most energetically favorable locations and converting selected grid-point energies into alignment-independent descriptors. By using a training set of 143 diverse chiral compounds, an optimal PLS model requiring seven components was chosen by using the cross-validation method resulting in a correlation coefficient R2= 0.88 and a cross validated correlation coefficient Q2= 0.85. An interpretation of the model is proposed based upon a visual inspection of the regression coefficient plots. From these plots, the influence of particular molecular features for selective binding of solutes was estimated and used to outline the chiral recognition sites in the Whelk-O1 receptor. The predictive power of our model has been estimated by means of an external data set emphasizing the suitability of the procedure also for predictive aims.

Theoretical analysis of the retention behavior of alcohols in gas chromatography

Quantitative structure-retention relationship (QSRR) models for the chromatographic (GC) retention times of alcohols on Superox 20M-diglycerol polarity stationary phase have been developed. Semi-empirical quantum chemical method (AM1) in MOPAC and Hartree-Fock (HF) method in Gaussian 98 implemented were employed to calculate a set of molecular descriptors of alcohols and ethyl acetate. Using multiple linear regression (MLR), we obtained the empirical functions with high correlation coefficient between retention times and quantum-chemical descriptors. The retention mechanism of alcohols of separation operating in the gas chromatogram was discussed. The results indicated that the QSRR models proposed were satisfactory.

Separation strategy for acidic chiral pharmaceuticals with capillary electrochromatography on polysaccharide stationary phases

The effect of five factors on the capillary electrochromatographic enantioseparation of acidic compounds was studied using an experimental design. The studied factors were pH, acetonitrile content in the mobile phase, temperature, buffer concentration, and applied voltage. These experiments allowed defining a generic separation strategy applicable on acidic compounds with chemical and structural diversity. The starting screening conditions consist of a 45 mM ammonium formate electrolyte at pH 2.9 mixed with 65% acetonitrile, an applied voltage of 15 kV, and a temperature of 25 degrees C. The screening phase occasionally can be followed by an optimization procedure. Evaluation of the proposed strategy pointed out that it allows achieving baseline resolution within a relatively short time when a beginning of separation is obtained at the starting conditions. This strategy revealed enantioselectivity for 11 compounds out of 15, of which 10 could be baseline-separated after the proposed optimization steps.

Data mining and enantiophore studies on chiral stationary phases used in HPLC separation

ChirBase database has been employed to mine the chemical structures of compounds resolved on common commercial chiral stationary phases (CSP). Different data sets were produced. The molecular fingerprint (enantiophore) strategy was then applied over these data sets. Enantiophores are identified by analyzing and mapping the three-dimensional common structural features shared by the ligand molecules imported from ChirBase. Such lists of encoded ligand enantiophores allowed us to generate 3D maps of the molecular interacting fragments, which are supposed to act reciprocally with each CSP. Results show that each CSP is combined with different preferential enantiophore counterparts in the ligand. These differences may be well related to the particular behavior of a given CSP to separate different families of compounds. As expected, supramolecular cellulosic or amylosic CSPs show generalist behavior by resolving a wide range of racemates. On the other hand, molecular CSPs based on a well-defined chiral receptor (such as Whelk-O1 or Chirobiotic T) appear to be more specific and thus specially adapted for more restrictive families of compounds. In addition, enantiophore analyses confirm that the supramolecular CSPs can combine multiple potential binding sites and so offer numerous enantioselective mechanisms toward a ligand. Inversely, on molecular CSPs, the ligand must respect strict geometrical constraints to make possible the chiral discrimination. Additional applications of the methodology are discussed as well.

Probability rule for chiral recognition

Molecular Chirality is of central interest in biological studies because enantiomeric compounds, while indistinguishable by most inanimate systems, show profoundly different properties in biochemical environments. Enantioselective separation methods, based on the differential recognition of two optical isomers by a chiral selector, have been amply documented. Also, great effort has been directed towards a theoretical understanding of the fundamental mechanisms underlying the chiral recognition process. Here we report a comprehensive data examination of enantio separation measurements for over 72000 chiral selector-select and pairs from the chiral selection compendium CHIRBASE. The distribution of alpha = k'(D)/k'(L) values was found to follow a power law, equivalent to an exponential decay for chiral differential free energies. This observation is experimentally relevant in terms of the number of different individual or combinatorial selectors that need to be screened in order to observe alpha values higher than a preset minimum. A string model for enantiorecognition (SMED) formalism is proposed to account for this observation on the basis of an extended Ogston three-point interaction model. Partially overlapping molecular interaction domains are analyzed in terms of a string complementarity model for ligand-receptor complementarity. The results suggest that chiral selection statistics may be interpreted in terms of more general concepts related to biomolecular recognition.

Modeling the chromatographic enantioseparation of aryl- and hetarylcarbinols on ULMO, a brush-type chiral stationary phase, by 3D-QSAR techniques

The enantiomers of a series of chiral aryl- and hetaryl-carbinols were separated on a chiral stationary phase (ULMO) based on (S,S)-3,5-dinitrobenzoylated 1,2-diphenylethane-1,2-diamine. In all cases, the homochiral analogs of (S)-1-phenylethanol (mostly also S-enantiomers) were retained more. Log alpha values were modeled with the aid of the 3D-QSAR techniques CoMFA (comparative molecular field analysis) and CoMSIA (comparative molecular similarity indices analysis) as well as a descriptor based on normal mode eigenvalues (EVA). Partial least-squares analysis with two (CoMSIA) or three (CoMFA, EVA) latent variables on a set of 22 training analytes gives cross-validated correlation coefficients q(2) = 0.85-0.91 and conventional correlation coefficients r(2) = 0.94-0.99. The quantitative structure-enantioselective retention relationships derived thereby were used to predict the separation factors of the test set, containing also hetaryl-carbinols, e.g., furan and thiophene analogs, with good accuracy.

Prediction of internal standards in reversed-phase liquid chromatography. III. Evaluation of an alternative solvation parameter model to correlate and predict the retention of ionizable compounds

This paper describes the results of the evaluation of an alternative solvation parameter model for ionizable compounds. The new model is described as Log(k) = Int + rR2 + spi2(H) + asigmaalpha2(H) + bsigmabeta2(H) + mVx + U/(1 + V10 (+/-(pH-Pk))). The first six terms are the usual solvation parameter equation for neutral solutes, and the last term represents the contribution to retention from the ionization of solutes. Retention data obtained for 30 solutes in acetonitrile/aqueous buffer mobile phases are used to evaluate the capability of the function using different pH/pK scales. Because the function is not linear, nonlinear least-squares analysis is used to perform the data processing. It is concluded that the model function describes similarly the retention of ionizable compounds to the literature model without the need to accurately measure the mobile phase pH and solute's pK. Accordingly, the function simplifies the application of linear solvation energy relationships (LSERs) to ionizable compounds, and allows us to easily predict their retention for chromatographic optimization.

Application of the artificial neural network in quantitative structure-gradient elution retention relationship of phenylthiocarbamyl amino acids derivatives

Quantitative structure-retention relationship(QSRR) method was used to model reversed-phase high-performance liquid chromatography (RP-HPLC) separation of 18 selected amino acids. Retention data for phenylthiocarbamyl (PTC) amino acids derivatives were obtained using gradient elution on ODS column with mobile phase of varying acetonitrile, acetate buffer and containing 0.5 ml/l of triethylamine (TEA). Molecular structure of each amino acid was encoded with 36 calculated molecular descriptors. The correlation between the molecular descriptors and the retention time of the compounds in the calibration set was established using the genetic neural network method. A genetic algorithm (GA) was used to select important molecular descriptors and supervised artificial neural network (ANN) was used to correlate mobile phase composition and selected descriptors with the experimentally derived retention times. Retention time values were used as the network's output and calculated molecular descriptors and mobile phase composition as the inputs. The best model with five input descriptors was chosen, and the significance of the selected descriptors for amino acid separation was examined. Results confirmed the dominant role of the organic modifier in such chromatographic systems in addition to lipophilicity (log P) and molecular size and shape (topological indices) of investigated solutes.