Computer-assisted prediction of gas chromatographic retention times of polychlorinated biphenyls

Machine Learning-Assisted Identification and Quantification of Hydroxylated Metabolites of Polychlorinated Biphenyls in Animal Samples

Laboratory studies of the disposition and toxicity of hydroxylated polychlorinated biphenyl (OH-PCB) metabolites are challenging because authentic analytical standards for most unknown OH-PCBs are not available. To assist with the characterization of these OH-PCBs (as methylated derivatives), we developed machine learning-based models with multiple linear regression (MLR) or random forest regression (RFR) to predict the relative retention times (RRT) and MS/MS responses of methoxylated (MeO-)PCBs on a gas chromatograph-tandem mass spectrometry system. The final MLR model estimated the retention times of MeO-PCBs with a mean absolute error of 0.55 min (n = 121). The similarity coefficients cos θ between the predicted (by RFR model) and experimental MS/MS data of MeO-PCBs were >0.95 for 92% of observations (n = 96). The levels of MeO-PCBs quantified with the predicted MS/MS response factors approximated the experimental values within a 2-fold difference for 85% of observations and 3-fold differences for all observations (n = 89). Subsequently, these model predictions were used to assist with the identification of OH-PCB 95 or OH-PCB 28 metabolites in mouse feces or liver by suggesting candidate ranking information for identifying the metabolite isomers. Thus, predicted retention and MS/MS response data can assist in identifying unknown OH-PCBs.

Potential of topological descriptors to model the retention of polychlorinated biphenyls in different gas chromatography stationary phases, including ionic liquid-based columns

The aim of this study was to develop a statistical model based on a set of intuitive topological descriptors that will help to determine the influence of the polychlorinated biphenyls (PCBs) structural features on the chromatographic behavior of these analytes in a variety of gas chromatographic stationary phases, including the highly polar ionic liquid (IL)-based SLB-IL76 and SLB-IL60 columns. The model was developed using the stepwise multiple linear regression method, and constructed through several levels of increasing complexity to make evident the relative influence of the selected descriptors. The proposed model was easy to implement and provided similar satisfactory results irrespective of the dependent variables used (i.e., retention index or retention time) or the chromatographic conditions applied (i.e., pseudo-isotherm and programmed temperature) for IL-based phases. The model also allowed the correct prediction of the elution order of selected PCBs in these and other less polar phases evaluated (i.e., SW-10, DB-17, ZB-5 and HT-8). To our knowledge, this is the first models based on topological descriptors described in the literature that provided a satisfactory fitting of the PCB behavior in IL-based phases. Our results indicated that the mechanism governing the chromatographic separation of PCBs in these highly polar columns showed significant differences compared with those observed in other less polar stationary phases.

Characterisation of Gas-Chromatographic Poly(Siloxane) Stationary Phases by Theoretical Molecular Descriptors and Prediction of McReynolds Constants

Retention in gas–liquid chromatography is mainly governed by the extent of intermolecular interactions between the solute and the stationary phase. While molecular descriptors of computational origin are commonly used to encode the effect of the solute structure in quantitative structure–retention relationship (QSRR) approaches, characterisation of stationary phases is historically based on empirical scales, the McReynolds system of phase constants being one of the most popular. In this work, poly(siloxane) stationary phases, which occupy a dominant position in modern gas–liquid chromatography, were characterised by theoretical molecular descriptors. With this aim, the first five McReynolds constants of 29 columns were modelled by multilinear regression (MLR) coupled with genetic algorithm (GA) variable selection applied to the molecular descriptors provided by software Dragon. The generalisation ability of the established GA-MLR models, evaluated by both external prediction and repeated calibration/evaluation splitting, was better than that reported in analogous studies regarding nonpolymeric (molecular) stationary phases. Principal component analysis on the significant molecular descriptors allowed to classify the poly(siloxanes) according to their chemical composition and partitioning properties. Development of QSRR-based models combining molecular descriptors of both solutes and stationary phases, which will be applied to transfer retention data among different columns, is in progress.

Novel Dechlorane Analogues and Possible Sources in Peregrine Falcon Eggs and Shark Livers from the Western North Atlantic Regions

During the investigation of dechlorane-related chemicals in North American wildlife, two unknown polychlorinated compounds (referred to as U1 and U2) were discovered. After extensive sample cleanup, structural information on U1 and U2 was characterized by gas chromatography (GC) coupled with single quadrupole mass spectrometer (MS) or GC-quadrupole time-of-flight (QToF) MS. Mass spectral evidence suggests that both U1 and U2 are structurally related to Dechlorane 603 (Dec603; C₁₇H₈Cl₁₂), an analogue of the chlorinated flame retardant Dechlorane Plus. From the results we suspect U1 (C₁₇H₉Cl₁₁) to be a monohydro analogue of Dec603 (i.e., one chlorine atom in Dec603 is replaced by a hydrogen atom). U1 may be formed via the degradation of Dec603's stereoisomers or present as an impurity in commercial Dec603 products. Mass spectral characterization of U2 (C₁₇H₇OCl₁₁) suggests it is a carbonylic derivative of Dec603, likely formed via metabolic transformation of Dec603 or its photoisomer. Semiquantitative measurement revealed that U1 and U2 were present at estimated median concentrations of 49 ng/g lipid weight (lw) and 59 ng/g lw in peregrine falcon ( Falco peregrinus) eggs, from the mid-Atlantic region of the United States, and 4.6 and 3.0 ng/g lw in shortfin mako shark ( Isurus oxyrinchus) livers from the western North Atlantic Ocean, respectively. Our results demonstrate the occurrence of these two novel Dec603-related chemicals in both terrestrial and aquatic ecosystems.

Predicting gas chromatography relative retention times for polychlorinated biphenyls using chlorine substitution pattern contribution method

Various quantitative structure retention relationships have been published in an effort to understand and predict chromatographic retention times. This work presents a chlorine substitution pattern contribution (Cl-SPC) model for relative retention times (RRT) of polychlorinated biphenyls (PCBs), using 27 sets of previously published gas chromatography RRT data. The Cl-SPC model calculates the contribution factors (βk) for each of 19 chlorine substitution "patterns" (such as 2-, 2,4-, 2,3,6-, 2,3,4,5,6-, etc.) using multiple linear regression (MLR). The 27 separate MLRs had R(2) values ranging from 0.961 to 1.000; the average absolute errors were 0.55% for the training sets and 0.95% for the test sets. Cross-validation of the model was carried out by splitting each data set into training and test sets for groupings based on nine PCB congener mixes commercialized by AccuStandard. No weakening of the model performance was observed when the size of data set used to develop the model was decreased from 209 to 39 congeners. In addition to the separate models, a single mixed model was fit combining all 27 data sets. The estimated random effects, which reflect the impact of GC configuration and operational conditions on RRTs, are minor compared with the fixed effects estimated for the βk values. The major advantages of the Cl-SPC model are its unmatched simplicity and equally excellent robustness when compared with other quantitative structure retention relationship models.

Investigation of retention behavior of polychlorinated biphenyl congeners on 18 different HRGC columns using molecular surface average local ionization energy descriptors

In this paper, based on the general interaction properties function (GIPF) family descriptors computed at the B3LYP/6-31G* level in Gaussian98 software, a significant quantitative structure-retention relationship (QSRR) models for the high resolution gas chromatographic relative retention time (HRGC-RRT) of all PCB congeners on 18 different HRGC capillary columns were constructed by using multiple linear regression (MLR) analysis, following the guidelines for development and validation of QSRR models. By means of the elimination selection stepwise regression algorithms, the molecular surface average local ionization energy was selected as one-parameter univariate linear regression to develop a QSRR model for prediction of GC-RRT of PCBs on each stationary phase. The accuracy of all developed models was confirmed using different types of internal and external procedures. A successful interpretation of the complex relationship between HRGC-RRTs of PCBs and the chemical structures was achieved by QSRR.

Semi-Empirical Topological Method for Prediction of the Relative Retention Time of Polychlorinated Biphenyl Congeners on 18 Different HR GC Columns

High resolution gas chromatographic relative retention time (HRGC-RRT) models were developed to predict relative retention times of the 209 individual polychlorinated biphenyls (PCBs) congeners. To estimate and predict the HRGC-RRT values of all PCBs on 18 different stationary phases, a multiple linear regression equation of the form RRT = a(o) + a(1) (no. o-Cl) + a(2) (no. m-Cl) + a(3) (no. p-Cl) + a(4) (V(M) or S(M)) was used. Molecular descriptors in the models included the number of ortho-, meta-, and para-chlorine substituents (no. o-Cl, m-Cl and p-Cl, respectively), the semi-empirically calculated molecular volume (V(M)), and the molecular surface area (S(M)). By means of the final variable selection method, four optimal semi-empirical descriptors were selected to develop a QSRR model for the prediction of RRT in PCBs with a correlation coefficient between 0.9272 and 0.9928 and a leave-one-out cross-validation correlation coefficient between 0.9230 and 0.9924 on each stationary phase. The root mean squares errors over different 18 stationary phases are within the range of 0.0108-0.0335. The accuracy of all the developed models were investigated using cross-validation leave-one-out (LOO), Y-randomization, external validation through an odd-even number and division of the entire data set into training and test sets. ELECTRONIC SUPPLEMENTARY MATERIAL: The online version of this article (doi:10.1365/s10337-010-1696-5) contains supplementary material, which is available to authorized users.

Prediction of gas chromatography/electron capture detector retention times of chlorinated pesticides, herbicides, and organohalides by multivariate chemometrics methods

A quantitative structure-retention relationship (QSRR) study, has been carried out on the gas chromatograph/electron capture detector (GC/ECD) system retention times (t(R)s) of 38 diverse chlorinated pesticides, herbicides, and organohalides by using molecular structural descriptors. Modeling of retention times of these compounds as a function of the theoretically derived descriptors was established by multiple linear regression (MLR) and partial least squares (PLS) regression. The stepwise regression using SPSS was used for the selection of the variables that resulted in the best-fitted models. Appropriate models with low standard errors and high correlation coefficients were obtained. Three types of molecular descriptors including electronic, steric and thermodynamic were used to develop a quantitative relationship between the retention times and structural properties. MLR and PLS analysis has been carried out to derive the best QSRR models. After variables selection, MLR and PLS methods used with leave-one-out cross validation for building the regression models. The predictive quality of the QSRR models were tested for an external prediction set of 12 compounds randomly chosen from 38 compounds. The PLS regression method was used to model the structure-retention relationships, more accurately. However, the results surprisingly showed more or less the same quality for MLR and PLS modeling according to squared regression coefficients R2 which were 0.951 and 0.948 for MLR and PLS, respectively.

Prediction of chromatographic relative retention time of polychlorinated biphenyls from the molecular electronegativity distance vector

Using the molecular electronegativity distance vector (MEDV) descriptors derived directly from the molecular topological structures, the gas chromatographic relative retention times (RRTs) of 209 polychlorinated biphenyls (PCBs) on the SE-54 stationary phase were predicted. A five-variable regression equation with the correlation coefficient of 0.9964 and the root mean square errors of 0.0152 was developed. The descriptors included in the equation represent degree of chlorination (nCl), nonortho index (Ino), and interactions between three pairs of atom types, i.e., atom groups -C= and -C=, -C= and >C=, -C= and -Cl. It has been proved that the retention times of all 209 PCB congeners can be accurately predicted as long as there are more than 50 calibration compounds. In the same way, the MEDV descriptors are also used to develop the five- or six-variable models of RRTs of PCBs on other 18 stationary phases and the correlation coefficients in both modeling stage and LOO cross-validation step are not lower than 0.99 except two models.

Predicting gas chromatographic retention times for the 209 polybrominated diphenyl ether congeners

A gas chromatographic relative retention time (GC-RRT) model was developed to predict retention times of the 209 individual polybrominated diphenyl ether (PBDE) congeners. Using the available 46 PBDE standards with mono- to deca-bromination, a multiple linear regression equation of the form RRT = b0 +b1 (no. o-Br) + b2(no. m-Br) +b3 (no. p-Br) +b4(mu) +b5 (ln MW) was used to predict the RRTs of the remaining 163 PBDE congeners. Molecular descriptors in the model included the number of ortho-, meta-, and para-bromine substituents (no. o-Br, m-Br andp-Br, respectively), the semi-empirically calculated dipole moment (mu), and the natural logarithm of molecular weight (MW). A high level of predictability (R2 = 0.9972) was obtained for the model.

Development of a multiple-class high-resolution gas chromatographic relative retention time model for halogenated environmental contaminants

A predictive model for the relative gas chromatographic retention times (GC-RRTs) of the following nine classes of halogenated environmental contaminants was developed: polybrominated diphenyl ethers (PBDEs); polychlorinated diphenyl ethers (PCDEs); polychlorinated biphenyls (PCBs); polychlorinated naphthalenes (PCNs); polychlorinated dibenzo-p-dioxins (PCDDs); polychlorinated dibenzofurans (PCDFs); polybrominated dibenzo-p-dioxins (PBDDs); polybrominated dibenzofurans (PBDFs); and organochlorine pesticides. MOPAC calculated physicochemical properties and structural descriptors in the model include molecular weight, square root of the number of halogen substituents, ionization potential, dipole moment, and the number of ortho, meta, and para halogen substituents. Using these variables, individual models for each of the contaminant classes were combined into a multiple class model incorporating the GC-RRTs of the 375 compounds of interest. The individual and multiclass GC-RRT models had acceptable fits between observed and predicted GC-RRTs (r2 = 0.9741-0.9990 for PBDEs, PCDEs, PCBs, PCNs, PCDD/Fs, and PBDD/Fs; r2 = 0.9250 for pesticides; and r2 = 0.9631 for the multiclass model) over a wide range of retention times and molecular structures. The combined model was tested on known GC-RRTs of hydroxylated PCBs and chlorinated phenoxyphenols and provided satisfactory results, demonstrating the strength of the model in predicting GC-RRT windows for contaminant classes not used in constructing the model. Such models will be useful in predicting the GC retention characteristics of novel environmental contaminants and their degradation products, for which analytical standards may not be available.

Chromatographic and ionization properties of polybrominated diphenyl ethers using GC/high-resolution MS with metastable atom bombardment and electron impact ionization

The chromatographic and ionization properties of 35 polybrominated diphenyl ether (PBDE) congeners were investigated using GC/HRMS with metastable atom bombardment (MAB) and electron impact (EI) ionization. A multiple linear regression model based on bromine substitution patterns and MOPAC calculated physical properties was developed to predict relative GC retention times of individual PBDE congeners. Although five different sources of metastable rare gas atoms (He, N2, Ar, Xe, and Kr) were investigated with MAB ionization, only MAB-N2 provided adequate ionization efficiency and predictability. Because of reduced background noise to the MS detector, MAB-N2 had a lower limit of detection for tetra- and penta-BDEs than EI, despite having a lower sensitivity. Using MAB-N2, the molecular ion was always the base peak, with little fragmentation taking place. Conversely, using EI ionization, the [M - nBr]+ peak (where n = 1-4, depending on the number of Br substituents) was the dominant ion for all PBDE congeners. Multiple linear regression models representing the molecular ion response of PBDE congeners analyzed by GC/ HRMS with MAB-N2 and EI ionization were also developed using the number and type of Br substituents and ionization potentials. A significantly higher level of predictability was obtained for the MAB-N2 response model than for EI.

Optimization of artificial neural networks used for retention modelling in ion chromatography

The aim of this work is the development of an artificial neural network model, which can be generalized and used in a variety of applications for retention modelling in ion chromatography. Influences of eluent flow-rate and concentration of eluent anion (OH-) on separation of seven inorganic anions (fluoride, chloride, nitrite, sulfate, bromide, nitrate, and phosphate) were investigated. Parallel prediction of retention times of seven inorganic anions by using one artificial neural network was applied. MATLAB Neural Networks ToolBox was not adequate for application to retention modelling in this particular case. Therefore the authors adopted it for retention modelling by programming in MATLAB metalanguage. The following routines were written; the division of experimental data set on training and test set; selection of data for training and test set; Dixon's outlier test; retraining procedure routine; calculations of relative error. A three-layer feed forward neural network trained with a Levenberg-Marquardt batch error back propagation algorithm has been used to model ion chromatographic retention mechanisms. The advantage of applied batch training methodology is the significant increase in speed of calculation of algorithms in comparison with delta rule training methodology. The technique of experimental data selection for training set was used allowing improvement of artificial neural network prediction power. Experimental design space was divided into 8-32 subspaces depending on number of experimental data points used for training set. The number of hidden layer nodes, the number of iteration steps and the number of experimental data points used for training set were optimized. This study presents the very fast (300 iteration steps) and very accurate (relative error of 0.88%) retention model, obtained by using a small amount of experimental data (16 experimental data points in training set). This indicates that the method of choice for retention modelling in ion chromatography is the artificial neural network.

Three-dimensional quantitative structure-property relationship (3D-QSPR) models for prediction of thermodynamic properties of polychlorinated biphenyls (PCBs): enthalpy of vaporization

Three-dimensional Quantitative Structure-Property Relationship (QSPR) models have been derived using Comparative Molecular Field Analysis (CoMFA) to correlate the vaporization enthalpies of a representative set of polychlorinated biphenyls (PCBs) at 298.15 K with their CoMFA-calculated physicochemical properties. Various alignment schemes, such as inertial, as is, and atom fit, were employed in this study. The CoMFA models were also developed using different partial charge formalisms, namely, electrostatic potential (ESP) charges and Gasteiger-Marsili (GM) charges. The most predictive model for vaporization enthalpy (Delta(vap)H(m)(298.15 K)), with atom fit alignment and Gasteiger-Marsili charges, yielded r2 values 0.852 (cross-validated) and 0.996 (conventional). The vaporization enthalpies of PCBs increased with the number of chlorine atoms and were found to be larger for the meta- and para-substituted isomers. This model was used to predict Delta(vap)H(m)(298.15 K) of the entire set of 209 PCB congeners.

Characterization of phosphorus-containing gas chromatographic stationary phases by linear solvation energy relationships

Linear solvation energy relationships allow the prediction of a variety of solubility interactions based on a set of descriptors found in the following equation: [equation: see text]. SP refers to an intrinsic thermodynamic property that can be found experimentally for a series of solutes. Phases containing phosphate, phosphite and phosphine functional groups were studied in this work. Coefficients obtained during this work, as well as those available for previously characterized phases, were correlated with molecular structural descriptors. When effects of non-phosphorus functional groups are estimated and subtracted out, hydrogen bond acceptor capability, a1, shows a positive trend when correlated with percent functional group. Correlation of the dipolarity/polarizability coefficient, s, with calculated atomic polarizability shows stationary phases group according to like functional groups. A similar correlation with dipole moment gives a trend of increasing dipole as s1 increases. Further quantitative structure-solubility relationship work is planned to better describe the contributions of inner shell and valence electrons to the chemical and physical properties of these compounds.

Three-dimensional quantitative structure--property relationship (3D-QSPR) models for prediction of thermodynamic properties of polychlorinated biphenyls (PCBs): enthalpy of sublimation

Three-dimensional quantitative structure--property relationship (3D-QSPR) models have been constructed using comparative molecular field analysis (CoMFA) to correlate the sublimation enthalpies at 298.15 K of a series of polychlorinated biphenyls (PCBs) with their CoMFA-calculated physicochemical properties. Various alignment schemes, such as atom fit, as is, and inertial were employed in this study. Separate CoMFA models were developed using different partial charge formalisms, namely, electrostatic potential (ESP) and Gasteiger-Marsili (GM) charges. Among the four different CoMFA models constructed for sublimation enthalpy (Delta(sub)H(m)(298.15 K)), the model that combined atom fit alignment and ESP charges yielded the greatest self-consistency (r(2) = 0.976) and internal predictive ability (r(cv)(2) = 0.750). This CoMFA model was used to predict Delta(sub)H(m)(298.15 K) of PCBs for which the corresponding experimental values are unavailable in the literature.

Quantitative structure-retention relationships for gas chromatographic retention indices of alkylbenzenes with molecular graph descriptors

Quantitative structure-retention relationships (QSRR) represent statistical models that quantify the connection between the molecular structure and the chromatographic retention indices of organic compounds, allowing the prediction of retention indices of novel, not yet synthesized compounds, solely from their structural descriptors. Using multiple linear regression, QSRR models for the gas chromatographic Kováts retention indices of 129 alkylbenzenes are generated using molecular graph descriptors. The correlational ability of structural descriptors computed from 10 molecular matrices is investigated, showing that the novel reciprocal matrices give numerical indices with improved correlational ability. A QSRR equation with 5 graph descriptors gives the best calibration and prediction results, demonstrating the usefulness of the molecular graph descriptors in modeling chromatographic retention parameters. The sequential orthogonalization of descriptors suggests simpler QSRR models by eliminating redundant structural information.

Quantitative relationship between chromatographic retentions and molecular structures of polychlorinated dibenzo-p-dioxins (PCDDs)

A new approach to study the quantitative relationships between chromatographic retentions and molecular structures of polychlorinated dibenzo-p-dioxins (PCDDs) is described. The retention equations of PCDDs log k' = A + B/T in gas chromatography (GC) are used to evaluate the properties of the regression coefficients A and B, which have been widely accepted as highly reliable chromatographic retentions. The quantitative relationships between the A, B values and the molecular structures are found. The molecular descriptors given for the first time in this article are very effective. As a result, the regression equations are derived with correlation coefficients greater than 0.9995. The A, B values of PCDDs with no standards available have been predicted according to these relationships. They are very useful in chromatographic identification. The retention times of all PCDDs can be conveniently predicted at any temperature program. Compared with the data obtained from the relevant experiments, the results of prediction are very accurate.

Quantitative structure-retention relationships in doping control

Regression equation modelling was used for the correlation of gas chromatographic relative retention times tRR of anabolic steroids, stimulants and narcotics with their molecular characteristics in order to create a model for the prediction of tRR values of unanalyzed molecules. Predicting chromatographic retention parameters is one of the main goals of the quantitative structure-retention relationships (QSRR) methodology. To be performed, QSRR studies require two tools; a methodology for the extraction of the structural characteristics and a statistical program for the correlation of these characteristics with the chromatographic data.

Polychlorinated biphenyls in the environment

This review surveys the problems arising from the release of PCBs into the environment from the point of view of the analytical chemist. These problems are very complex and interdependent and so it is essential to recognize their mutual links rather than to separate one problem from another (sources of contamination, fate in the environment, toxic properties and particular capabilities, limitations and purposes of analytical methods). Prominent attention should be paid in the future to congener-specific analyses of "toxic" congeners using high-resolution gas chromatography and to toxicity-assessing biological methods.