Prediction Of Retention Factors Research Articles

Influence of mixed salts on retention behavior of model proteins in cation exchange chromatography

Mobile phase composition is an important factor for a further improvement of ion exchange chromatography steps of proteins. In this work, the effects of mixed salts on the retention factors of the two model proteins lysozyme (LYZ) and bovine serum albumin (BSA) in cation exchange chromatography (CEC) were investigated and compared to effects previously observed in hydrophobic interaction chromatography (HIC). The model equation describing the effects in HIC was adjusted for linear gradient elution experiments in CEC. The investigated salts were sodium chloride, sodium sulfate, ammonium chloride and ammonium sulfate. By varying binary salt mixtures as well as using pure salts, model parameters were determined. The normalized root mean square error (NRMSE) of the predicted retention factors for the calibration runs was 4.1% for BSA and 3.1% for LYZ. Additional validation experiments proved the ability of the model to describe and predict retention behavior of the proteins for further salt compositions. Hereby, the NRMSE values for BSA and LYZ were 2.0% and 1.5%, respectively. While the retention factors of LYZ changed linearly with the salt composition, non-linearities in the impact of the anion composition were found for BSA. This was contributed to an overlay of a synergetic salt effect on a protein-specific effect by sulfate on BSA with non-specific effects of the ions for CEC. However, the impact of the synergetic effects on protein separation is lower for CEC than for HIC, as mixed salts do not increase the separation of these proteins. The best salt composition for separating BSA and LYZ is pure ammonium sulfate. Thus, synergetic salt effects can also occur in CEC, but they have a lower impact than in HIC.

Deep Q-learning for the selection of optimal isocratic scouting runs in liquid chromatography

An important challenge in chromatography is the development of adequate separation methods. Accurate retention models can significantly simplify and expedite the development of adequate separation methods for complex mixtures. The purpose of this study was to introduce reinforcement learning to chromatographic method development, by training a double deep Q-learning algorithm to select optimal isocratic scouting runs to generate accurate retention models. These scouting runs were fit to the Neue-Kuss retention model, which was then used to predict retention factors both under isocratic and gradient conditions. The quality of these predictions was compared to experimental data points, by computing a mean relative percentage error (MRPE) between the predicted and actual retention factors. By providing the reinforcement learning algorithm with a reward whenever the scouting runs led to accurate retention models and a penalty when the analysis time of a selected scouting run was too high (> 1h); it was hypothesized that the reinforcement learning algorithm should by time learn to select good scouting runs for compounds displaying a variety of characteristics. The reinforcement learning algorithm developed in this work was first trained on simulated data, and then evaluated on experimental data for 57 small molecules – each run at 10 different fractions of organic modifier (0.05 to 0.90) and four different linear gradients. The results showed that the MRPE of these retention models (3.77% for isocratic runs and 1.93% for gradient runs), mostly obtained via 3 isocratic scouting runs for each compound, were comparable in performance to retention models obtained by fitting the Neue-Kuss model to all (10) available isocratic datapoints (3.26% for isocratic runs and 4.97% for gradient runs) and retention models obtained via a “chromatographer's selection” of three scouting runs (3.86% for isocratic runs and 6.66% for gradient runs). It was therefore concluded that the reinforcement learning algorithm learned to select optimal scouting runs for retention modeling, by selecting 3 (out of 10) isocratic scouting runs per compound, that were informative enough to successfully capture the retention behavior of each compound.

Quantitative structure retention relationship modeling as potential tool in chromatographic determination of stability constants and thermodynamic parameters of β-cyclodextrin complexation process

When cyclodextrins (CDs) are used in chromatography analytes’ retention time is decreased with an increase in concentration of CD in the mobile phase. Thus complex stability constants can be determined from the change in retention time of the ligand molecule upon complexation. Since the preceding approach implies extensive and time-consuming HPLC experiments, the goal of this research was to investigate the possibility of using in silico prediction tools instead. Quantitative structure–retention relationship (QSRR) model previously developed to explain the retention behavior of risperidone, olanzapine and their structurally related impurities in β-CD modified HPLC system was applied to predict retention factor under different chromatographic conditions within the examined domains. Predicted retention factors were further used for calculation of stability constants and important thermodynamic parameters, namely standard Gibbs free energy, standard molar enthalpy and entropy, contributing to inclusion phenomenon. Unexpected prolonged retention with an increase in β-CD concentration was observed, in contrast to the employed chromatographic theory used for the calculation of the stability constants. Consequently, it led to failure in stability constants and thermodynamic parameters calculation for almost all analytes when acetonitrile content was 20% (v/v) across the investigated pH range. Moreover, ionization of investigated analytes and free stationary phase silanol groups are pH dependent, leading to minimization of secondary interactions if free silanol groups are non-ionized at pH lower than 3. In order to prove accuracy of predicted retention factors, HPLC verification experiments were performed and good agreement between predicted and experimental values was obtained, confirming the applicability of proposed in-silico tool. However, the obtained results opened some novel questions and revealed that chromatographic method is not overall applicable in calculation of stability constants and thermodynamic parameters indicating the complexity of β-CD modified systems.

Generic approach to the method development of intact protein separations using hydrophobic interaction chromatography.

We describe a liquid chromatography method development approach for the separation of intact proteins using hydrophobic interaction chromatography. First, protein retention was determined as function of the salt concentration by isocratic measurements and modeled using linear regression. The error between measured and predicted retention factors was studied while varying gradient time (between 15 and 120min) and gradient starting conditions, and ranged between 2 and 15%. To reduce the time needed to develop optimized gradient methods for hydrophobic interaction chromatography separations, retention-time estimations were also assessed based on two gradient scouting runs, resulting in significantly improved retention-time predictions (average error<2.5%) when varying gradient time. When starting the scouting gradient at lower salt concentrations (stronger eluent), retention time prediction became inaccurate in contrast to predictions based on isocratic runs. Application of three scouting runs and a nonlinear model, incorporating the effects of gradient duration and mobile-phase composition at the start of the gradient, provides accurate results (improved fitting compared to the linear solvent-strength model) with an average error of 1.0% and maximum deviation of -8.3%. Finally, gradient scouting runs and retention-time modeling have been applied for the optimization of a critical-pair protein isoform separation encountered in a biotechnological sample.

PREDICTION OF HPLC RETENTION FACTOR OF POTENTIAL ANTITUBERCULOTICS BY QSRR

Derivatives of 3-phenyl-2H-1,3-benzoxazine-2,4(3H)-dione (PBOD) have become important mainly as perspective antituberculotic drugs. Quantitative structure-retention relationship (QSRR) is used for predicting the HPLC retention factor of this group of compounds and optimal chromatographic conditions appropriate for this purpose are selected. Among many molecular properties utilizable as the QSRR descriptors, mainly, in silico variables are advantageous as they closely characterize the HPLC retention of the PBOD molecule. Additionally, they are available without a need of the compound synthesis, which is important in the first stage of development of the potential drug. Artificial neural networks (ANN) were successfully used as the basic modeling QSRR tools because their regression outputs allow a direct prediction of retention factors for different combinations of the stationary and mobile phases. To prevent overfitting in ANN calculations, a distinctive elimination step, reducing about 50% of initial variables, was necessary to assure a desirable QSRR performance. Thoroughly tuned ANN models enabled a very accurate prediction of the PBOD retention factors when using C18 stationary phase and acetonitrile/water or methanol/water mobile phase.

Study of Retention in Micellar Liquid Chromatography on a C18 Column on the Basis of Linear Solvation Energy Relationships

In this study, 8 solutes (aniline, caffeine, p-cresol, ethyl benzene, methylparaben, phenol, pyridine, and toluene) have been tested in terms of linear solvation energy relationships (LSER). Several micellar liquid chromatography (MLC) systems using cationic surfactant cetyltrimethylammonium bromide (CTAB) and a mixture of water with (methanol, n-propanol, and n-butanol) modifiers were characterized using the LSER solvation parameter model. The effects of the surfactant and modifier concentration on the retention in MLC were discussed. LSER model had demonstrated high potential to predict retention factors with high squared correlation coefficients (<TEX>$r^2$</TEX> > 0.99). A comparison of predicted and experimental retention factors suggests that LSER formalism is able to reproduce adequately the experimental retention factors of the solutes studied in the different experimental conditions investigated. This model is a helpful tool to understand the solute-surfactant interactions and evaluate the retention characteristic of micellar liquid chromatography.

Chromatography in Silico, Quantitative Analysis of Retention of Aromatic Acid Derivatives

A quantitative analysis of the retention of aromatic acid derivatives in reversed-phase liquid chromatography (RPLC) is conducted using a molecular mechanics calculation in the CAChe program. The molecular interaction energy value is calculated by subtracting the energy value of the complex from the sum of energy values of a model phase and an analyte. Several model phases are constructed, and the feasibility of applying the method to a variety of compounds is examined based on improving the contact surface area and the capability of computer software and hardware. Interaction energy values are calculated for both molecular and ionic forms. The predicted retention factors of partially ionized acids obtained using a combination of dissociation constants correlated well with the values measured by RPLC with pH-controlled eluents.

Insights into the retention mechanism on an octadecylsiloxane-bonded silica stationary phase (HyPURITY C18) in reversed-phase liquid chromatography

Plots of the retention factor against mobile phase composition were used to organize a varied group of solutes into three categories according to their retention mechanism on an octadecylsiloxane-bonded silica stationary phase HyPURITY C18 with methanol–water and acetonitrile–water mobile phase compositions containing 10–70% (v/v) organic solvent. The solutes in category 1 could be fit to a general retention model, Eq. (2), and exhibited normal retention behavior for the full composition range. The solutes in category 2 exhibited normal retention behavior at high organic solvent composition with a discontinuity at low organic solvent compositions. The solutes in category 3 exhibited a pronounced step or plateau in the middle region of the retention plots with a retention mechanism similar to category 1 solutes at mobile phase compositions after the discontinuity and a different retention mechanism before the discontinuity. Selecting solutes and appropriate composition ranges from the three categories where a single retention mechanism was operative allowed modeling of the experimental retention factors using the solvation parameter model. These models were then used to predict retention factors for solutes not included in the models. The overwhelming number of residual values [log k (experimental) − log k (model predicted)] were negative and could be explained by contributions from steric repulsion, defined as the inability of the solute to insert itself fully into the stationary phase because of its bulkiness (i.e., volume and/or shape). Steric repulsion is shown to strongly depend on the mobile phase composition and was more significant for mobile phases with a low volume fraction of organic solvent in general and for mobile phases containing methanol rather than acetonitrile. For mobile phases containing less than about 20 % (v/v) organic solvent the mobile phase was unable to completely wet the stationary phase resulting in a significant change in the phase ratio and for acetonitrile (but less so methanol) changes in the solvation environment indicated by a discontinuity in the system maps.

Contribution of Steric Repulsion to Retention on an Octadecylsiloxane-Bonded Silica Stationary Phase in Reversed-Phase Liquid Chromatography

Plots of the retention factor against mobile phase composition were used to organize a varied group of solutes into three categories according to their retention mechanism on an octadecylsilioxane-bonded silica stationary phase, Ascentis TM C18, with acetonitrile-water and methanol-water mobile phase compositions containing 10–70% (v/v) organic solvent. The solutes in category 1 could be fit to a general retention model, Eq. (1), and exhibited normal retention behavior for the full composition range. The solutes in category 2 exhibited normal retention behavior at high organic solvent compositions with a discontinuity at low organic solvent mobile phase compositions. The solutes in category 3 exhibited a pronounced step or plateau in the middle region of the retention plots with a retention mechanism similar to category 1 at mobile phase compositions after the discontinuity and a different retention mechanism before the discontinuity. Selecting solutes and appropriate composition ranges from the three categories where a single retention mechanism was operative allowed modeling of the experimental retention factors using the solvation parameter model. These models were then used to predict retention factors for solutes excluded from the models. The overwhelming number of residual values, here defined as the difference between experimental and model predicted retention factors for the excluded solutes, could be explained by contributions from steric repulsion. The latter defined as the inability of solutes to fully insert themselves into the solvated stationary phase because of their size or conformation. Steric repulsion resulted in a systematic reduction in retention compared with predicted values for the fully inserted solute. The bonding density of the stationary phase; the type and composition of the mobile phase; and the size, conformation, type and number of functional groups on the solute are shown to affect the contribution of steric repulsion to the retention mechanism.

Prediction of Retention Factors in Supercritical Fluid Chromatography Using Artificial Neural Network

In this study, a quantitative structure-property relationship technique has been used for the prediction of retention factors for some organic compounds in supercritical fluid chromatography using an artificial neural network. The best descriptors that appear in this model are the number of single bonds, the number of double bonds, and the hydrophilic factor. These descriptors were used as inputs for a generated artificial neural network. This network has a 3 : 3 : 1 topology that was trained using a back-propagation algorithm. The cross-validation method was used to evaluate the predictive power of the generated network. The results obtained by artificial neural networks were compared with the experimental values as well as with those obtained using the multiple linear regression technique. Comparison of these results shows the ability of the artificial neural network model to predict retention factors.

Chromatography In Silico for Basic Drugs

The retention factors in liquid chromatography can be quantitatively predicted from the molecular interaction energy calculated by molecular mechanics, using a model phase and nitrogen‐containing compounds including basic drugs. The new system can also predict retention factors of ionized compounds. However, an old system using log P could not predict the retention of ionized compounds. The correlation coefficient between the molecular interaction energy and retention factors of both molecular and ionized forms was better than that between log P and retention factors. The addition of pKa to molecular interaction energy could enable one to predict retention factors of partially ionized compounds.

Separation of small peptides by electrochromatography on silica-based reversed phases and hydrophobic anion exchange phases

Peptide separations are regarded as a promising application of capillary electrochromatography (CEC) and, at the same time, a suitable model to elucidate its mixed separation mechanism when charged analytes are involved. In this paper, studies on the separation of small peptides (2-4 amino acids) on a Spherisorb octadecyl silane (ODS) phase at acidic pH and on a strong anion exchange (SAX)/C18 mixed mode phase at weakly basic pH are reported. For the ODS phase a comparison of CEC, capillary zone electrophoresis (CZE) and high-performance liquid chromatography (HPLC) under identical buffer/eluent conditions is presented. The predicted retention factors for CEC under the assumption of simple superposition of HPLC retention and CZE migration matched the measured results for the peptides that had small retention factors in HPLC. For both types of stationary phases, a variation of the acetonitrile content in the mobile phase led to a wide range of retention factors, including negative values when co-electroosmotic migration was dominant. Though both the ODS and the SAX/C18 phase offer unique advantages, the SCX/C18 phase at pH 9 provides more flexibility to alter separation selectivity for the selected peptides.

Chromatography in silico, basic concept in reversed-phase liquid chromatography

Basic phenomena in reversed-phase liquid chromatography have been quantitatively analyzed using a computational chemical calculation. Pyridine interacted with an ionized silica surface under neutral conditions. Alkyl-chain length affected the contact surface area with an analyte. Steric hindrance was demonstrated using a model graphitic carbon phase and unsaturated alkenes. Quantitative structure-retention relationships in reversed-phase liquid chromatography were demonstrated for phenolic compounds and acidic and basic drugs. The correlations between predicted and measured retention factors were satisfactory. Dissociation constants were derived from the atom partial charge and used to predict retention factors of partially ionized compounds.

Modelling retention in liquid chromatography of polyphenolic acids: Prediction of solvent composition and pH of the mobile phase

The retention behavior of a series of polyphenolic acids has been modeled in water–acetonitrile mobile phases, in order to optimize the chromatographic resolution between these compounds. The retention model is based on the relationships between retention factors of solutes with the pH and solvatochromic parameter of the mobile phase ( E T N ) , and between the dissociation constants of the compounds and the mole fraction of organic modifier. The retention of these ionizable compounds has been studied at three levels of acetonitrile percentage (10, 20 and 30%), and at six different pH values ranging between 3.5 and 7. Three different fitting procedures have been applied: model I consists of a nonlinear regression to all data points, whereas models II and III use only four and two data points, respectively, taken from the extreme pH / E T N values. The results obtained from these models indicate a good agreement between real and predicted retention factors for each compound, with errors about 5% for model I and about 10% for models II and III. The results of model III are of particular interest, because of only two experiments for each compound are needed to describe the retention behavior in a relatively large pH and percentage of acetonitrile intervals; this is especially valuable in the optimization of the mobile phase for chromatographic separation of several compounds.

Chromatography in silica, quantitative analysis of retention mechanisms of benzoic acid derivatives

A quantitative analysis of the retention of benzoic acid derivatives in reversed-phase liquid chromatography was achieved using a molecular mechanics calculation in the CAChe program. Interaction energy values were calculated for both molecular and ionic forms. The predicted retention factors of partially ionized acids obtained using a combination of dissociation constants well correlated with the values measured by reversed-phase liquid chromatography with pH-controlled eluents. The molecular interaction energy value was calculated by subtracting the energy value of the complex from the sum of energy values of a model-phase and an analyte.

Monomeric and polymeric anionic gemini surfactants and mixed surfactant systems in micellar electrokinetic chromatography. Part II: Characterization of chemical selectivity using two linear solvation energy relationship models

Sodium di(undecenyl) tartarate monomer (SDUT), a vesicle-forming amphiphilic compound possessing two hydrophilic carboxylate headgroups and two hydrophobic undecenyl chains, was prepared and polymerized to form a polymeric vesicle (i.e., poly-SDUT). The anionic surfactants of SDUT and poly-SDUT (carboxylate head group) and sodium dodecyl sulfate, SDS (sulfate head groups) as well as mixed surfactant systems (SDS/SDUT, SDS/poly-SDUT, and SDUT/poly-SDUT) were applied as pseudostationary phases in micellar electrokinetic chromatography (MEKC). Two linear solvation energy relationship (LSER) models, i.e., solvatochromic and solvation parameter models, were successfully applied to investigate the effect of the type and composition of pseudostationary phases on the retention mechanism and selectivity in MEKC. The solvatochromic and solvation parameter models were used to help understand the fundamental nature of the solute-pseudostationary phase interactions and to characterize the properties of the pseudostationary phases (e.g., solute size and hydrogen bond-accepting ability for all pseudostationary phases). The solute types were found to have a significant effect on the LSER system coefficients and on the predicted retention factors. Although both LSER models provide the same information, the solvation parameter model is found to provide much better results both statistically and chemically than the solvatochromic model.

Predictions of micelle-water partition coefficients and retention in micellar electrokinetic chromatography from solute structure. 2. Fragmental constant approach.

The fragmental constant approach (FCA) was used to calculate water-sodium dodecyl sulfate (SDS) micelle partition coefficients, K(mw), for uncharged solutes from their structure. Subsequently, the availability of K(mw) values allows prediction of retention factor, k, in micellar electrokinetic chromatography (MEKC) using the simple relationship k = K(mw)phi, where phi is the phase ratio. The FCA model describes a micelle-water partition coefficient as the sum of the partition coefficients of the constituent atomic/molecular fragments, measured by fragmental constant values, f (i), as well as correction factors to account for various "intramolecular effects" that cause deviations from the predicted partition coefficients as, log K(mw) = sum(n)(i=1)aif i+sum(m)(i=1)kiCm. The fragmental constants for a set of 41 fragments were determined using a training set of 229 aromatic solutes and 198 aliphatic compounds. The K(mw) of the aromatic compounds in the training set were determined by MEKC, while the K(mw) of the aliphatic solutes were estimated using the linear solvation energy relationship (LSER) for the SDS micelles. The fragments consisted of both aromatic fragments (i.e., directly attached to an aromatic ring) and aliphatic fragments. The FCA predictions agree nicely with the observed and LSER partition coefficient values, even for complex molecular structures such as beta-blocker drugs. The results show the great potential of the FCA for a priori prediction of retention behavior in MEKC from solute structure.

Simulation of chromatography of phenolic compounds with a computational chemical method

An ab initio simulation of reversed-phase liquid chromatography for phenolic compounds was achieved based on molecular interaction energy values calculated using molecular mechanics calculations (MM2) of the CAChe program. The precision of the predicted retention factors from the molecular interaction energy values was equivalent to the predicted retention factors based on octanol–water partition coefficients (log P) calculated using the molecular orbital package (MOPAC). The prediction of retention factors of phenolic compounds in reversed-phase liquid chromatography in a given pH eluent was possible using the predicted dissociation constant (p K a) from the atomic partial charge without a chemical experiment if the organic modifier effect was known.

Quantitative structure–retention relationships of phenolic compounds without Hammett’s equations

Retention times of phenolic compounds in a given pH eluent in reversed-phase liquid chromatography were predicted from dissociation constants derived from atomic partial charges and log P-values calculated by a computational chemical method. The precision of the calculation of atomic partial charges by AM1 and PM3 methods of MOPAC™ was evaluated. The atomic partial charges obtained by AM1 were the more acceptable. The atomic partial charges obtained from the hydrogen of the hydroxyl group included an ortho-effect, therefore an ortho-effect was added to the predicted p K a values. The precision of predicted retention factors obtained using predicted p K a values was similar to that using reference p K a values.

COMPUTATIONAL CHEMICAL PREDICTION OF THE RETENTION FACTOR OF AROMATIC ACIDS

ABSTRACT The retention factors of aromatic acids were predicted from their log P and pKa values in reversed-phase liquid chromatography in eluent at a given pH. The log P value, a modified CAChe-log P ™ and Vlog P ™, was calculated by computational chemical methods. The pKa values in this study were derived from partial charges of atoms by computational chemical methods instead of Hammett's equations. The precision of predicted retention factors was dependent on the accuracy of pKa values.