Applying Quantitative Structure–Activity Relationship (QSAR) Methodology for Modeling Postmortem Redistribution of Benzodiazepines and Tricyclic Antidepressants

Journal of Analytical Toxicology 2014;38:242 –248 doi:10.1093/jat/bku025 Advance Access publication March 28, 2014 Article Applying Quantitative Structure–Activity Relationship (QSAR) Methodology for Modeling Postmortem Redistribution of Benzodiazepines and Tricyclic Antidepressants Constantinos Giaginis1*, Anna Tsantili-Kakoulidou2 and Stamatios Theocharis3 1 Department of Food Science and Nutrition, School of Environment, University of the Aegean, Mitropoliti Ioakeim 2 Street, Myrina, Lemnos GR 81400, Greece, 2Department of Pharmaceutical Chemistry, School of Pharmacy, University of Athens, Panepistimiopolis, Zografou, Athens 157 71, Greece, and 3First Department of Pathology, Medical School, University of Athens, Athens, Greece *Author to whom correspondence should be addressed. Email: Postmortem redistribution (PMR) constitutes a multifaceted process, which complicates the interpretation of drug concentrations by forensic toxicologists. The present study aimed to apply quantitative structure –activity relationship (QSAR) analysis for modeling PMR data of structurally related drugs, 10 benzodiazepines and 10 tricyclic antidepressants. For benzodiazepines, an adequate QSAR model was obtained (R 2 5 0.98, Q 2 5 0.88, RMSEE 5 0.12), in which energy, ionization and molecular size exerted significant impact. For tricyclic antidepressants, an adequate QSAR model with slightly inferior statistics (R 2 5 0.95, Q 2 5 0.87, RMSEE 5 0.29) was established after exclusion of maprotiline, in which energy parameters, basicity character and lipophilicity exerted significant contribution. Thus, QSAR analysis could be used as a complementary tool to provide an informative illustration of the contributing molecular, physicochemical and structural properties in PMR process. However, the complexity, nonstatic and time-dependent nature of PMR endpoints raises serious concerns whether QSAR methodology could predict the degree of redistribution, highlighting the need for animal-derived PMR data. Introduction The interpretation of drug concentrations in postmortem period constitutes one of the most difficult aspects in the field of forensic toxicology. This is mainly ascribed to the fact that drug concentrations obtained from postmortem samples do not necessarily reflect the blood concentration at the time of death due to variations according to the sampling site and the interval between death and sampling (1 – 5). These site- and timedependent variations have been attributed to a multifaceted process occurring after death, referred to as postmortem redistribution (PMR) (6, 7). The high complexity of PMR process is amplified by the fact that it does not exclusively refer to fatal poisoning, but it may also reflect antemortem physical or biochemical alterations. Many drugs are sequestered antemortem in organs qualified as drug reservoirs. Hollow organs with high concentrating power, such as different parts of the gastrointestinal tract or viscera, liver, lungs and heart, can function as drug reservoirs (6 –8). Redistribution from these organs can occur by diffusion through blood vessels and/or transparietal diffusion toward surrounding organs (6 –8). The competing processes of diffusion from drug reservoirs, disruption of cellular membranes and putrefaction may lead to alterations in drug concentrations between sites and sampling intervals (9, 10). Drugs biotransformation by metabolizing enzymes, such as cytochrome P450 monooxygenases and/or uridine diphosphate-glucuronosyltransferases, could also be triggered at an early stage of postmortem period (11). Consequently, PMR of drugs may complicate the interpretation of the analytical results in forensic toxicology and has emergently been considered as ‘a toxicological nightmare’ (10, 12). Quantitative structure – activity relationship (QSAR) methodology is a useful tool for analyzing in a systematic way the information incorporated in the chemical structure of compounds in relation to the available biological data (13, 14). This methodology has widely been accepted to evaluate and predict the activity of drugs against a therapeutic target, as well as the toxicity risk assessment of drugs or chemicals, supporting the reduction, refinement and/or replacement of experimental studies (15 – 17). Currently, relevant software packages are available, which permit the calculation of a large number of physicochemical, molecular and structural descriptors, facilitating their elaboration in QSARs as an alternative strategy for the establishment of predictive models (14–16). In the last few years, multivariate data analysis (MVDA), a projection method to latent variables, the principal components, which are linear combination of the original descriptors, has widely been applied to QSAR modeling (18, 19). MVDA has been considered as a powerful statistical tool, which can treat a large number of interrelated descriptors, exploiting the maximum information encoded within them and separating regularities from noise (18, 19). Moreover, MVDA permits the elaboration of descriptors and response variables, in which the high accuracy of computational and experimental values is not an emergent requisite in contrast to conventional statistical methods (18, 19). Physicochemical properties are considered to play a crucial role in pharmacokinetics, mainly in absorption and distribution, as well as in the pharmacodynamic and toxicological profile of drugs (20, 21). Among them, lipophilicity, ionization and volume of distribution were reported to affect the ability of drugs to redistribute across the tissue barriers during postmortem period (22 – 24). In a previous work by our group, we effectively applied QSAR methodology to model PMR of structurally diverse drugs (25). The derived QSAR model provided an informative illustration of the contributing molecular, physicochemical and structural properties of drugs in the PMR process (25). However, this model presented moderate predictive power and limited applicability, which was ascribed to the high complexity of the PMR process and the structurally diversity of the data set (25). In view of above considerations, the present study aimed to develop more specific QSAR models using two structurally homogeneous data sets, benzodiazepines and tricyclic antidepressants, which attract special attention in the field of forensic toxicology. # The Author 2014. Published by Oxford University Press. All rights reserved. For Permissions, please email: Methods and materials Data set The data set consisted of two chemical classes of compounds, 10 benzodiazepines and 10 tricyclic antidepressants. The extent of PMR expressed by the central : peripheral concentration ratio (C : P ratio) was taken from the literature (22). Descriptors Structural specific constitutional descriptors were derived manually. Structural specific constitutional descriptors include the total number of rings (nRings), the number of aromatic rings (nArC6), the number (...truncated)