Maria G. Kouskoura

Modelling by partial least squares the relationship between the HPLC mobile phases and analytes on phenyl column

Twenty-five descriptors and 61 structurally different analytes have been used on a partial least squares (PLS) to latent structure technique in order to study chromatographically their interaction mechanism on a phenyl column. According to the model, 240 different retention times of the analytes, expressed as Y variable (log k), at different % MeOH mobile-phase concentrations have been correlated with their theoretical most important structural or molecular descriptors. The goodness-of-fit was estimated by the coefficient of multiple determinations r(2) (0.919), and the root mean square error of estimation (RMSEE=0.1283) values with a predictive ability (Q(2)) of 0.901. The model was further validated using cross-validation (CV), validated by 20 response permutations r(2) (0.0, 0.0146), Q(2) (0.0, -0.136) and validated by external prediction. The contribution of certain mechanism interactions between the analytes, the mobile phase and the column, proportional or counterbalancing is also studied. Trying to evaluate the influence on Y of every variable in a PLS model, VIP (variables importance in the projection) plot provides evidence that lipophilicity (expressed as Log D, Log P), polarizability, refractivity and the eluting power of the mobile phase are dominant in the retention mechanism on a phenyl column.

Using the partial least squares method to model the electrospray ionization response produced by small pharmaceutical molecules in positive mode.

RATIONALE The signal intensity in electrospray ionization mass spectrometry (ESI-MS) positive mode is affected by parameters that are related to the physicochemical properties and structural features of a molecule. Accordingly, the combined interactions of an analyte and the mobile phase used is still an area that demands further clarifications since there is no general pattern regarding the nature of a molecule and the mechanism by which vapor-phase ions are produced. METHODS A multivariate analysis method, such as Partial Least Squares (PLS), provides the opportunity to correlate the effect of a large number of parameters interpreting this complex procedure with the use of appropriate mathematical algorithms. This work involves the development of models containing up to 84 X variables which characterize the analytes studied (99) focusing on their positive or negative effect on the vapor-phase ion formation process. These descriptors are correlated with the signal response of the positively charged analyte ions which corresponds to the Y variable. RESULTS The results showed that parameters referring or directly related to the ionization percentage of basic or acidic groups of an analyte can be used to determine the signal response on positive ESI-MS mode. Structural characteristics, polar surface area, lipophilicity, the ability of analytes to acts as hydrogen bond donors or acceptors, water solubility, density of a solid, surface tension of the substance and the number of free rotatable bonds are descriptors which are of secondary importance, still they cannot be considered negligible. CONCLUSIONS The models derived are proved to be reliable for the investigation of such mechanisms, with a small number of components and good linearity (R(2) >83%, Q(2) >70%).

Negative electrospray ionization mode in mass spectrometry: A new perspective via modeling

Electrospray ionization technique is used for production of gas phase ions without fragmentation and is considered as one of the most sensitive analytical methods for structural characterization of molecules. Nonetheless, the determination of some parameters (physicochemical properties or structural features) that may enhance the signal response especially in the negative ion mode has not yet been clarified. The present work is an attempt to correlate the signal response behavior of 110 compounds used as probes, with their characteristics (molecular descriptors, X variables). In order to quantify this phenomenon, Partial Least Squares which is a software capable of performing linear multivariate analysis was applied. The models derived explore the positive or negative effect of 49 X variables on the signal response of each analyte, expressed as Y variable. The process of gas phase ions formation was verified by both flow injection and column analysis. The models derived are proven reliable for the study of such mechanisms, with small number of components and good linearity (R(2)>83%, Q(2)>70%). The present study showed that parameters as pKa, ionization percentage of the analyte, PSA, HBA, COOH, water solubility and surface tension of a solid are affecting ion formation. At the same time, slight differentiations of the influence of certain parameters were observed on column injection analysis due to the chromatographic delay of some analytes.

HILIC Chromatography An Insight on the Retention Mechanism

Hydrophilic interaction chromatography (HILIC) could be characterized as a complex chromatographic system that involves multiple mechanisms. These are partitioning as well as polar and ionic interactions. Among several HILIC columns, ZIC-HILIC can be used to separate small organic ionic compounds. The presence of both positive and negative charge on the stationary phase may facilitate separations of both anionic and cationic analytes. Based on the Partial Least Squares methodology, an attempt to clarify the mechanism on this column revealed that the forces dominating are mainly determined by structural features. Consequently, the physicochemical properties which are related to the analytes’ structure may heighten or attenuate the process. Ionic interactions are stronger for analytes containing moieties with basic properties since the interaction with the sulfonyl group is facilitated. The partition mechanism is prevailing for those analytes that are not sufficiently ionized at the experimental conditions (mobile phase pH 3 and 6.5) and for analytes that can create halogen bonds. Moreover, the stagnant water layer on the silica bed enhances the retention of water soluble compounds due to the increased hydrophilic interactions.

Decoding the signal response of steroids in electrospray ionization mode (ESI-MS)

Electrospray ionization (ESI) is predominant among soft ionization techniques since it is considered as the method of choice for coupling liquid chromatography with mass spectrometry (LC-MS). Despite the progress which has been achieved in the ion formation theory, the research community keep their interest in the parameters affecting the increase in the responsiveness of the signal. This particular problem is becoming more complex when the analytes studied are compounds not having characteristic moieties, which are responsible for a molecules ionization (carboxylic or amine groups). The present study attempts to decode the signal intensity by correlating it with a series of structural features and physicochemical properties corresponding to 30 steroids. These molecules share a common basic structure with only small differences in the substitution while they do not contain any basic or acidic group (pKbasic 10.6). The correlation and evaluation of the significance of the parameters causing an increase or decrease in the signal response was achieved using multivariate analysis via the Partial Least Squares methodology (PLS). Moreover, the PLS models that were developed could be used as predictive tools of the signal intensity for unknown substances.

A new descriptor via bio-mimetic chromatography and modeling for the blood brain barrier (Part II)

Graphical abstract Figure. No Caption available. HighlightsModeling the ability of 51 CNS drugs to permeate the BBB via chromatography.Development of a predictive model by Partial Least Squares methodology.Introduction a new parameter describing the drugs behavior in the context of modeling drugs behavior.Exploration of the effect of physicochemical and pharmacokinetic properties. Abstract Within the context of drug design methodology for the central nervous system (CNS), a predictive model which can shorten the process of finding new candidate drugs was developed. Therefore, the retention time of 51 molecules which are clinically established to enter the blood brain barrier (BBB), were recorded on two HPLC columns. For this purpose, a lipophilic butyl (C4) stationary phase was used to simulate the behavior of a drug regarding BBB permeability and a zwitterionic‐HILIC to simulate blood. The results were plotted as Y variables on two Partial Least Squares (PLS) models, while 25 specific physicochemical properties (significant for lipid bilayers BBB permeation or blood) were used as X descriptors. Both models can be utilized to predict the drugability of a new molecule avoiding needless animal experiments, as well as time and material consuming syntheses. The developed models were validated (R2 ≥ 0.90, Q2 ≥ 0.83), and based on the results specific variables were proved to be significant for the studied phenomenon. Additionally, a new factor symbolized as MT was introduced. MT incorporated the experimental results and it was calculated by the fraction of the sum of the retention time of the drug on the two columns (tr(butyl) + tr(HILIC)) divided by the molecular volume (Vm) of each analyte. This new descriptor was used as an equivalent to the logarithm of BBB permeability (logBB) and may indicate the ability of a new molecule to act as a candidate drug able to enter the BBB. Comprehending the extend of contribution of several molecular attributes to the in vivo distribution of a drug may enlighten the knowledge on pharmacokinetics and clinical variation, and enable scientists to design more efficient drug molecules.

Elucidation of the retention mechanism on a reverse‐phase cyano column by modeling

Partial least squares and quantitative structure-retention relationship models have been used mainly to explain and then to predict the retention mechanism on a cyanopropyl high-performance liquid chromatography column. Developing and applying the models involves studying the chromatographic behavior of 100 probes. Characterization of the probes took place under optimized isocratic conditions at variable proportions of two mobile phase mixtures. Retention time was correlated with numerous physicochemical properties and structural features of the probes. The goodness-of-fit for both models was estimated by the coefficient of multiple determinations, while the prediction of a test set was achieved by the root mean square error of prediction. The contribution of the descriptors in partial least squares is confirmed by the information derived from the variable importance in the projection and loadings plots, while a quantitative structure-retention relationship reflects the behavior model. In both cases, the descriptors determining the retention mechanism are lipophilicity, solubility in water, molecular volume and the presence of -COOH and/or condensed rings. Such techniques are proven useful tools for visualizing, exploring, and modeling the complex interactions between solutes and the mobile and stationary phase while at the same time this information can be quantified.

A Complementary Study of Mechanisms and Behaviors in Chromatography via Modeling.

Evolution in preparation of chromatographic columns has created the need for studying and evaluating them with the use of smart software. This research is an attempt to compare the retention mechanism between two stationary phases (butyl and phenyl) with the use of multivariate analysis for a large number of probes. Partial least squares has the ability to spot either major or minor differences in the chromatographic behavior of probes, with regard to changes in the stationary or mobile phases. The models developed refer to a total of 108 miscellaneous chemical compounds, described by 63 X variables (physicochemical properties and structural features) and one Y variable (retention time). The results showed that in both columns and mobile phases (40% methanol or 40% acetonitrile) the retention of an analyte is mainly affected by its lipophilicity, molar volume, and refractivity, which tend to cause delayed elution. On the contrary, solubility in water, polar surface area, and hydrogen bond donor or acceptor properties promote faster elution. The most important difference proved to be the effect of the presence of the carboxylic group and the solubility that affected the retention in a similar way in both columns but not with both mobile phases.