Barun Bhhatarai

Prediction of Aqueous Solubility, Vapor Pressure and Critical Micelle Concentration for Aquatic Partitioning of Perfluorinated Chemicals

The majority of perfluorinated chemicals (PFCs) are of increasing risk to biota and environment due to their physicochemical stability, wide transport in the environment and difficulty in biodegradation. It is necessary to identify and prioritize these harmful PFCs and to characterize their physicochemical properties that govern the solubility, distribution and fate of these chemicals in an aquatic ecosystem. Therefore, available experimental data (10-35 compounds) of three important properties: aqueous solubility (AqS), vapor pressure (VP) and critical micelle concentration (CMC) on per- and polyfluorinated compounds were collected for quantitative structure-property relationship (QSPR) modeling. Simple and robust models based on theoretical molecular descriptors were developed and externally validated for predictivity. Model predictions on selected PFCs were compared with available experimental data and other published in silico predictions. The structural applicability domains (AD) of the models were verified on a bigger data set of 221 compounds. The predicted properties of the chemicals that are within the AD, are reliable, and they help to reduce the wide data gap that exists. Moreover, the predictions of AqS, VP, and CMC of most common PFCs were evaluated to understand the aquatic partitioning and to derive a relation with the available experimental data of bioconcentration factor (BCF).

Modelling physico-chemical properties of (benzo)triazoles, and screening for environmental partitioning.

(Benzo)triazoles are distributed throughout the environment, mainly in water compartments, because of their wide use in industry where they are employed in pharmaceutical, agricultural and deicing products. They are hazardous chemicals that adversely affect humans and other non-target species, and are on the list of substances of very high concern (SVHC) in the new European regulation of chemicals - REACH (Registration, Evaluation, Authorization and Restriction of Chemical substances). Thus there is a vital need for further investigations to understand the behavior of these compounds in biota and the environment. In such a scenario, physico-chemical properties like aqueous solubility, hydrophobicity, vapor pressure and melting point can be useful. However, the limited availability and the high cost of lab testing prevents the acquisition of necessary experimental data that industry must submit for the registration of these chemicals. In such cases a preliminary analysis can be made using Quantitative Structure-Property Relationships (QSPR) models. For such an analysis, we propose Multiple Linear Regression (MLR) models based on theoretical molecular descriptors selected by Genetic Algorithm (GA). Training and prediction sets were prepared a priori by splitting the available experimental data, which were then used to derive statistically robust and predictive (both internally and externally) models. These models, after verification of their structural applicability domain (AD), were used to predict the properties of a total of 351 compounds, including those in the REACH preregistration list. Finally, Principal Component Analysis was applied to the predictions to rank the environmental partitioning properties (relevant for leaching and volatility) of new and untested (benzo)triazoles within the AD of each model. Our study using this approach highlighted compounds dangerous for the aquatic compartment. Similar analyses using predictions obtained by the EPI Suite and VCCLAB tools are also compared and discussed in this paper.

Per- and Polyfluoro Toxicity (LC50 Inhalation) Study in Rat and Mouse Using QSAR Modeling

Fully or partially fluorinated compounds, known as per- and polyfluorinated chemicals are widely distributed in the environment and released because of their use in different household and industrial products. Few of these long chain per- and polyfluorinated chemicals are classified as emerging pollutants, and their environmental and toxicological effects are unveiled in the literature. This has diverted the production of long chain compounds, considered as more toxic, to short chains, but concerns regarding the toxicity of both types of per- and polyfluorinated chemicals are alarming. There are few experimental data available on the environmental behavior and toxicity of these compounds, and moreover, toxicity profiles are found to be different for the types of animals and species used. Quantitative structure-activity relationship (QSAR) is applied to a combination of short and long chain per- and polyfluorinated chemicals, for the first time, to model and predict the toxicity on two species of rodents, rat (Rattus) and mouse (Mus), by modeling inhalation (LC(50)) data. Multiple linear regression (MLR) models using the ordinary-least-squares (OLS) method, based on theoretical molecular descriptors selected by genetic algorithm (GA), were used for QSAR studies. Training and prediction sets were prepared a priori, and these sets were used to derive statistically robust and predictive (both internally and externally) models. The structural applicability domain (AD) of the model was verified on a larger set of per- and polyfluorinated chemicals retrieved from different databases and journals. The descriptors involved, the similarities, and the differences observed between models pertaining to the toxicity related to the two species are discussed. Chemometric methods such as principal component analysis (PCA) and multidimensional scaling (MDS) were used to select most toxic compounds from those within the AD of both models, which will be subjected to experimental tests under the EU project CADASTER.

CADASTER QSPR Models for Predictions of Melting and Boiling Points of Perfluorinated Chemicals

Quantitative structure property relationship (QSPR) studies on per‐ and polyfluorinated chemicals (PFCs) on melting point (MP) and boiling point (BP) are presented. The training and prediction chemicals used for developing and validating the models were selected from Syracuse PhysProp database and literatures. The available experimental data sets were split in two different ways: a) random selection on response value, and b) structural similarity verified by self‐organizing‐map (SOM), in order to propose reliable predictive models, developed only on the training sets and externally verified on the prediction sets. Individual linear and non‐linear approaches based models developed by different CADASTER partners on 0D‐2D Dragon descriptors, E‐state descriptors and fragment based descriptors as well as consensus model and their predictions are presented. In addition, the predictive performance of the developed models was verified on a blind external validation set (EV‐set) prepared using PERFORCE database on 15 MP and 25 BP data respectively. This database contains only long chain perfluoro‐alkylated chemicals, particularly monitored by regulatory agencies like US‐EPA and EU‐REACH. QSPR models with internal and external validation on two different external prediction/validation sets and study of applicability‐domain highlighting the robustness and high accuracy of the models are discussed. Finally, MPs for additional 303 PFCs and BPs for 271 PFCs were predicted for which experimental measurements are unknown.

Oral LD50 toxicity modeling and prediction of per- and polyfluorinated chemicals on rat and mouse

Quantitative structure-activity relationship (QSAR) analyses were performed using the LD50 oral toxicity data of per- and polyfluorinated chemicals (PFCs) on rodents: rat and mouse. PFCs are studied under the EU project CADASTER which uses the available experimental data for prediction and prioritization of toxic chemicals for risk assessment by using the in silico tools. The methodology presented here applies chemometrical analysis on the existing experimental data and predicts the toxicity of new compounds. QSAR analyses were performed on the available 58 mouse and 50 rat LD50 oral data using multiple linear regression (MLR) based on theoretical molecular descriptors selected by genetic algorithm (GA). Training and prediction sets were prepared a priori from available experimental datasets in terms of structure and response. These sets were used to derive statistically robust and predictive (both internally and externally) models. The structural applicability domain (AD) of the models were verified on 376 per- and polyfluorinated chemicals including those in REACH preregistration list. The rat and mouse endpoints were predicted by each model for the studied compounds, and finally 30 compounds, all perfluorinated, were prioritized as most important for experimental toxicity analysis under the project. In addition, cumulative study on compounds within the AD of all four models, including two earlier published models on LC50 rodent analysis was studied and the cumulative toxicity trend was observed using principal component analysis (PCA). The similarities and the differences observed in terms of descriptors and chemical/mechanistic meaning encoded by descriptors to prioritize the most toxic compounds are highlighted.

Evaluation of TOPKAT, Toxtree, and Derek Nexus in Silico Models for Ocular Irritation and Development of a Knowledge-Based Framework To Improve the Prediction of Severe Irritation

Assessment of ocular irritation is an essential component of any risk assessment. A number of (Q)SARs and expert systems have been developed and are described in the literature. Here, we focus on three in silico models (TOPKAT, BfR rulebase implemented in Toxtree, and Derek Nexus) and evaluate their performance using 1644 in-house and 123 European Centre for Toxicology and Ecotoxicology of Chemicals (ECETOC) compounds with existing in vivo ocular irritation classification data. Overall, the in silico models performed poorly. The best consensus predictions of severe ocular irritants were 52 and 65% for the in-house and ECETOC compounds, respectively. The prediction performance was improved by designing a knowledge-based chemical profiling framework that incorporated physicochemical properties and electrophilic reactivity mechanisms. The utility of the framework was assessed by applying it to the same test sets and three additional publicly available in vitro irritation data sets. The prediction of severe ocular irritants was improved to 73-77% if compounds were filtered on the basis of AlogP_MR (hydrophobicity with molar refractivity). The predictivity increased to 74-80% for compounds capable of preferentially undergoing hard electrophilic reactions, such as Schiff base formation and acylation. This research highlights the need for reliable ocular irritation models to be developed that take into account mechanisms of action and individual structural classes. It also demonstrates the value of profiling compounds with respect to their chemical reactivity and physicochemical properties that, in combination with existing models, results in better predictions for severe irritants.

Are Mechanistic and Statistical QSAR Approaches Really Different? MLR Studies on 158 Cycloalkyl-Pyranones

Two parallel approaches for quantitative structure‐activity relationships (QSAR) are predominant in literature, one guided by mechanistic methods (including read‐across) and another by the use of statistical methods. To bridge the gap between these two approaches and to verify their main differences, a comparative study of mechanistically relevant and statistically relevant QSAR models, developed on a case study of 158 cycloalkyl‐pyranones, biologically active on inhibition (Ki) of HIV protease, was performed. Firstly, Multiple Linear Regression (MLR) based models were developed starting from a limited amount of molecular descriptors which were widely proven to have mechanistic interpretation. Then robust and predictive MLR models were developed on the same set using two different statistical approaches unbiased of input descriptors. Development of models based on Statistical I method was guided by stepwise addition of descriptors while Genetic Algorithm based selection of descriptors was used for the Statistical II. Internal validation, the standard error of the estimate, and Fisher’s significance test were performed for both the statistical models. In addition, external validation was performed for Statistical II model, and Applicability Domain was verified as normally practiced in this approach. The relationships between the activity and the important descriptors selected in all the models were analyzed and compared. It is concluded that, despite the different type and number of input descriptors, and the applied descriptor selection tools or the algorithms used for developing the final model, the mechanistical and statistical approach are comparable to each other in terms of quality and also for mechanistic interpretability of modelling descriptors. Agreement can be observed between these two approaches and the better result could be a consensus prediction from both the models.

A QSAR study of HIV protease inhibitors using theoretical descriptors.

This paper reports the development of quantitative structure-activity relationship (QSAR) models for a set of 170 chemicals using mathematical descriptors which can be calculated directly from molecular structure without the input of any other experimental data. The calculated descriptors include topostructural (TS), topochemical (TC), and quantum chemical (QC). Because the situation is rank deficient i.e. the number of independent variables (descriptors) is larger than the number of compounds, three robust linear statistical modeling methods capable of handling such situations, viz., principal components regression (PCR), partial least square (PLS), and ridge regression (RR) were used for QSAR formulation. Results show that PLS and RR gave better q2 values as compared to the PCR method. Of the three classes of descriptors, the TC indices were the best predictors of anti-HIV activity and the QC indices were the least effective.

Evaluation of OASIS QSAR Models Using ToxCast™ in Vitro Estrogen and Androgen Receptor Binding Data and Application in an Integrated Endocrine Screening Approach.

Background: Integrative testing strategies (ITSs) for potential endocrine activity can use tiered in silico and in vitro models. Each component of an ITS should be thoroughly assessed. Objectives: We used the data from three in vitro ToxCast™ binding assays to assess OASIS, a quantitative structure-activity relationship (QSAR) platform covering both estrogen receptor (ER) and androgen receptor (AR) binding. For stronger binders (described here as AC50 < 1 μM), we also examined the relationship of QSAR predictions of ER or AR binding to the results from 18 ER and 10 AR transactivation assays, 72 ER-binding reference compounds, and the in vivo uterotrophic assay. Methods: NovaScreen binding assay data for ER (human, bovine, and mouse) and AR (human, chimpanzee, and rat) were used to assess the sensitivity, specificity, concordance, and applicability domain of two OASIS QSAR models. The binding strength relative to the QSAR-predicted binding strength was examined for the ER data. The relationship of QSAR predictions of binding to transactivation- and pathway-based assays, as well as to in vivo uterotrophic responses, was examined. Results: The QSAR models had both high sensitivity (> 75%) and specificity (> 86%) for ER as well as both high sensitivity (92–100%) and specificity (70–81%) for AR. For compounds within the domains of the ER and AR QSAR models that bound with AC50 < 1 μM, the QSAR models accurately predicted the binding for the parent compounds. The parent compounds were active in all transactivation assays where metabolism was incorporated and, except for those compounds known to require metabolism to manifest activity, all assay platforms where metabolism was not incorporated. Compounds in-domain and predicted to bind by the ER QSAR model that were positive in ToxCast™ ER binding at AC50 < 1 μM were active in the uterotrophic assay. Conclusions: We used the extensive ToxCast™ HTS binding data set to show that OASIS ER and AR QSAR models had high sensitivity and specificity when compounds were in-domain of the models. Based on this research, we recommend a tiered screening approach wherein a) QSAR is used to identify compounds in-domain of the ER or AR binding models and predicted to bind; b) those compounds are screened in vitro to assess binding potency; and c) the stronger binders (AC50 < 1 μM) are screened in vivo. This scheme prioritizes compounds for integrative testing and risk assessment. Importantly, compounds that are not in-domain, that are predicted either not to bind or to bind weakly, that are not active in in vitro, that require metabolism to manifest activity, or for which in vivo AR testing is in order, need to be assessed differently. Citation: Bhhatarai B, Wilson DM, Price PS, Marty S, Parks AK, Carney E. 2016. Evaluation of OASIS QSAR models using ToxCast™ in vitro estrogen and androgen receptor binding data and application in an integrated endocrine screening approach. Environ Health Perspect 124:1453–1461; http://dx.doi.org/10.1289/EHP184

Possible allosteric interactions of monoindazole-substituted P2 cyclic urea analogues with wild-type and mutant HIV-1 protease

Our ongoing efforts to understand the difference in the binding pattern of HIV-1 protease inhibitor (HIVPI) with the wild-type and mutant HIV-1 protease (HIVPR) and to provide mechanistic insight are continued further. We report here the results of a recent quantitative structure–activity relationship (QSAR) study on monoindazole-substituted P2 analogues of cyclic urea HIVPIs. The QSAR models revealed an inverted parabolic relationship between biological activity and calculated molar refractivity (CMR). That is, biological activity first decreases with increase in CMR and at a certain minimum point (inversion point) it suddenly changes and increases with further increase in CMR. CMR is a measure of volume-dependent-polarizability and is an indication of the polar interactions between ligand and receptor. The results seem to be best rationalized by larger molecules inducing a change in a receptor unit that allows for a new mode of interaction. Similar QSAR models were also observed for the biological activity of these molecules tested against a panel of mutant viruses including mutant strains with single amino acid substitution (I84V), double amino acid substitutions (I84V/V82F), and multiple amino acid changes corresponding to mutations observed in clinical isolates of patients treated with Ritonavir®. Interestingly the inversion points for these mutant strains were found larger than for wild-type. The subtle but significant difference in the inversion point indicates change in the shape and size of the binding pocket. Earlier QSAR studies have shown that the correlation of biological activity with an inverted parabola is an indicative of the ‘allosteric interaction’ of the ligands with the receptor. This report presents a detail analysis of these observations.