Latest advances in computational genotoxicity prediction

Use of In Silico Methods for Regulatory Toxicological Assessment of Pharmaceutical Impurities

The use of novel non-testing methodologies to support the toxicological assessment of drug impurities is having a growing impact in the regulatory framework for pharmaceutical development and marketed products. For DNA reactive (mutagenic) impurities specific recommendations for the use of in silico structure-based approaches (namely (Q)SAR methodologies) are provided in the ICH M7 guideline. In 2018 a draft reflection paper has been published by EMA addressing open issues in the qualification approach of non-genotoxic impurities (NGI) according to the ICH Q3A/Q3B guidelines, and proposing the use of alternative testing strategies, including TTC, (Q)SAR, read-across, and in vitro approaches, to gather impurity-specific safety information.In the present chapter we describe a workflow to perform the safety assessment of drug impurities based on non-testing in silico methodologies. The proposed approach consists of a stepwise decision scheme including three key phases: PHASE 1: assessment of bacterial mutagenicity and consequent classification of impurities according to ICH M7; PHASE 2: risk characterization of mutagenic impurities (Classes 1, 2 or 3); PHASE 3: qualification of non-mutagenic impurities (Classes 4 or 5). The proposed decision scheme offers the possibility to acquire impurity-specific data, also if testing is not feasible, and to decide on further in vitro testing, besides meeting 3R's principle.

Mode of action-based risk assessment of genotoxic carcinogens

The risk assessment of chemical carcinogens is one major task in toxicology. Even though exposure has been mitigated effectively during the last decades, low levels of carcinogenic substances in food and at the workplace are still present and often not completely avoidable. The distinction between genotoxic and non-genotoxic carcinogens has traditionally been regarded as particularly relevant for risk assessment, with the assumption of the existence of no-effect concentrations (threshold levels) in case of the latter group. In contrast, genotoxic carcinogens, their metabolic precursors and DNA reactive metabolites are considered to represent risk factors at all concentrations since even one or a few DNA lesions may in principle result in mutations and, thus, increase tumour risk. Within the current document, an updated risk evaluation for genotoxic carcinogens is proposed, based on mechanistic knowledge regarding the substance (group) under investigation, and taking into account recent improvements in analytical techniques used to quantify DNA lesions and mutations as well as "omics" approaches. Furthermore, wherever possible and appropriate, special attention is given to the integration of background levels of the same or comparable DNA lesions. Within part A, fundamental considerations highlight the terms hazard and risk with respect to DNA reactivity of genotoxic agents, as compared to non-genotoxic agents. Also, current methodologies used in genetic toxicology as well as in dosimetry of exposure are described. Special focus is given on the elucidation of modes of action (MOA) and on the relation between DNA damage and cancer risk. Part B addresses specific examples of genotoxic carcinogens, including those humans are exposed to exogenously and endogenously, such as formaldehyde, acetaldehyde and the corresponding alcohols as well as some alkylating agents, ethylene oxide, and acrylamide, but also examples resulting from exogenous sources like aflatoxin B1, allylalkoxybenzenes, 2-amino-3,8-dimethylimidazo[4,5-f] quinoxaline (MeIQx), benzo[a]pyrene and pyrrolizidine alkaloids. Additionally, special attention is given to some carcinogenic metal compounds, which are considered indirect genotoxins, by accelerating mutagenicity via interactions with the cellular response to DNA damage even at low exposure conditions. Part C finally encompasses conclusions and perspectives, suggesting a refined strategy for the assessment of the carcinogenic risk associated with an exposure to genotoxic compounds and addressing research needs.

In Silico Approaches in Predictive Genetic Toxicology

Genetic toxicology testing is a weight-of-evidence approach to identify and characterize chemical substances that can cause genetic modifications in somatic and/or germ cells. Prediction of genetic toxicology using computational tools is gaining more attention and preferred by regulatory authorities as an alternate safety assessment for in vivo or in vitro approaches. Due to the cost and time associated with experimental genetic toxicity tests, it is essential to develop more robust in silico methods to predict chemical genetic toxicity. A number of in silico genotoxicity predictive tools/models are developed based on the experimental data gathered over the years. These in silico tools are divided into statistical quantitative structure-activity relationships (QSAR)-based approaches and expert-based systems. This chapter covers the state of the art in silico toxicology approaches and standardized protocols, essential for conducting genetic toxicity predictions of chemicals. This chapter also highlights various parameters for the validation of the prediction results obtained from QSAR models.

Combining machine learning models of in vitro and in vivo bioassays improves rat carcinogenicity prediction

In vitro genotoxicity bioassays are cost-efficient methods of assessing potential carcinogens. However, many genotoxicity bioassays are inappropriate for detecting chemicals eliciting non-genotoxic mechanisms, such as tumour promotion, this necessitates the use of in vivo rodent carcinogenicity (IVRC) assays. In silico IVRC modelling could potentially address the low throughput and high cost of this assay. We aimed to develop and combine computational QSAR models of novel bioassays for the prediction of IVRC results and compare with existing software. QSAR models were generated from existing Ames (n = 6512), Syrian Hamster Embryonic (SHE, n = 410), ISSCAN rodent carcinogenicity (ISC, n = 834) and GreenScreen GADD45a-GFP (n = 1415) chemical datasets. These models mapped the molecular descriptors of each compound to their respective assay result using machine learning algorithms (adaboost, k-Nearest Neighbours, C.45 Decision Tree, Multilayer Perceptron, Random Forest). The best performing models were combined with k-Nearest Neighbours to create a cascade model for IVRC prediction. High QSAR model performance was observed from ten time 10-fold cross-validation with above 80% accuracy and 0.85 AUC for each assay dataset. The cascade model predicted rat carcinogenicity with 69.3% accuracy and 0.700 AUC. This study demonstrates the novelty of a combined approach for IVRC prediction, with higher performance than existing software.

In silico prediction of the mutagenicity of nitroaromatic compounds using a novel two-QSAR approach

Certain drugs are nitroaromatic compounds, which are potentially toxic. As such, it is of practical importance to assess and predict their mutagenic potency in the process of drug discovery. A classical quantitative structure-activity relationship (QSAR) model was developed using the linear partial least square (PLS) scheme to understand the underline mutagenic mechanism and a non-classical QSAR model was derived using the machine learning-based hierarchical support vector regression (HSVR) to predict the mutagenicity of nitroaromatic compounds based on a series of mutagenicity data (TA98-S9). It was observed that HSVR performed better than PLS as manifested by the predictions of the samples in the training set, test set, and outlier set as well as various statistical validations. A mock test designated to mimic real challenges also confirmed the better performance of HSVR. Furthermore, HSVR exhibited superiority in predictivity, generalization capabilities, consistent performance, and robustness when compared with various published predictive models. PLS, conversely, revealed some mechanistically interpretable relationships between descriptors and mutagenicity. Thus, this two-QSAR approach using the predictive HSVR and interpretable PLS models in a synergistic fashion can be adopted to facilitate drug discovery and development by designing safer drug candidates with nitroaromatic moiety.

Taking Advantage of Databases

Toxicity databases are a useful resource to support hazard and risk assessment. They are used to retrieve historical studies for compounds of interest and to support toxicity predictions where no data exists. Toxicity predictions are either based upon study results from similar chemicals or prediction models built from these databases.

The Consultancy Activity on In Silico Models for Genotoxic Prediction of Pharmaceutical Impurities

The toxicological assessment of DNA-reactive/mutagenic or clastogenic impurities plays an important role in the regulatory process for pharmaceuticals; in this context, in silico structure-based approaches are applied as primary tools for the evaluation of the mutagenic potential of the drug impurities. The general recommendations regarding such use of in silico methods are provided in the recent ICH M7 guideline stating that computational (in silico) toxicology assessment should be performed using two (Q)SAR prediction methodologies complementing each other: a statistical-based method and an expert rule-based method.Based on our consultant experience, we describe here a framework for in silico assessment of mutagenic potential of drug impurities. Two main applications of in silico methods are presented: (1) support and optimization of drug synthesis processes by providing early indication of potential genotoxic impurities and (2) regulatory evaluation of genotoxic potential of impurities in compliance with the ICH M7 guideline. Some critical case studies are also discussed.

Establishing best practise in the application of expert review of mutagenicity under ICH M7

The ICH M7 guidelines for the assessment and control of DNA reactive (mutagenic) impurities in pharmaceuticals allows for the consideration of in silico predictions in place of in vitro studies. This represents a significant advance in the acceptance of (Q)SAR models and has resulted from positive interactions between modellers, regulatory agencies and industry with a shared purpose of developing effective processes to minimise risk. This paper discusses key scientific principles that should be applied when evaluating in silico predictions with a focus on accuracy and scientific rigour that will support a consistent and practical route to regulatory submission.

An evaluation of in-house and off-the-shelf in silico models: implications on guidance for mutagenicity assessment

The evaluation of impurities for genotoxicity using in silico models are commonplace and have become accepted by regulatory agencies. Recently, the ICH M7 Step 4 guidance was published and requires two complementary models for genotoxicity assessments. Over the last ten years, many companies have developed their own internal genotoxicity models built using both public and in-house chemical structures and bacterial mutagenicity data. However, the proprietary nature of internal structures prevents sharing of data and the full OECD compliance of such models. This analysis investigated whether using in-house internal compounds for training models is needed and substantially impacts the results of in silico genotoxicity assessments, or whether using commercial-off-the-shelf (COTS) packages such as Derek Nexus or Leadscope provide adequate performance. We demonstrated that supplementation of COTS packages with a Support Vector Machine (SVM) QSAR model trained on combined in-house and public data does, in fact, improve coverage and accuracy, and reduces the number of compounds needing experimental assessment, i.e., the liability load. This result indicates that there is added value in models trained on both internal and public structures and incorporating such models as part of a consensus approach improves the overall evaluation. Lastly, we optimized an in silico consensus decision-making approach utilizing two COTS models and an internal (SVM) model to minimize false negatives.

Computational toxicology, friend or foe?

Computational approaches offer the attraction of being both fast and cheap to run being able to process thousands of chemical structures in a few minutes. As with all new technology, there is a tendency for these approaches to be hyped up and claims of reliability and performance may be exaggerated. So just how good are these computational methods?

(Q)SAR assessments of potentially mutagenic impurities: a regulatory perspective on the utility of expert knowledge and data submission

(Quantitative) structure activity relationship [(Q)SAR] modeling is the primary tool used to evaluate the mutagenic potential associated with drug impurities. General recommendations regarding the use of (Q)SAR in regulatory decision making have recently been provided in the ICH M7 guideline. Although (Q)SAR alone is capable of achieving reasonable sensitivity and specificity, reliance on a simple positive or negative prediction can be problematic. The key to improving (Q)SAR performance is to integrate supporting information, also referred to as expert knowledge, into the final conclusion. In the regulatory context, expert knowledge is intended to (1) maximize confidence in a (Q)SAR prediction, (2) provide rationale to supersede a positive or negative (Q)SAR prediction, or (3) provide a basis for assessing mutagenicity in absence of a (Q)SAR prediction. Expert knowledge is subjective and is associated with great variability in regards to content and quality. However, it is still a critical component of impurity evaluations and its utility is acknowledged in the ICH M7 guideline. The current paper discusses the use of expert knowledge to support regulatory decision making, describes case studies, and provides recommendations for reporting data from (Q)SAR evaluations.

Emerging approaches in predictive toxicology

Predictive toxicology plays an important role in the assessment of toxicity of chemicals and the drug development process. While there are several well-established in vitro and in vivo assays that are suitable for predictive toxicology, recent advances in high-throughput analytical technologies and model systems are expected to have a major impact on the field of predictive toxicology. This commentary provides an overview of the state of the current science and a brief discussion on future perspectives for the field of predictive toxicology for human toxicity. Computational models for predictive toxicology, needs for further refinement and obstacles to expand computational models to include additional classes of chemical compounds are highlighted. Functional and comparative genomics approaches in predictive toxicology are discussed with an emphasis on successful utilization of recently developed model systems for high-throughput analysis. The advantages of three-dimensional model systems and stem cells and their use in predictive toxicology testing are also described.

Hybrid in silico models for drug-induced liver injury using chemical descriptors and in vitro cell-imaging information

Drug-induced liver injury (DILI) is a major adverse drug reaction that accounts for one-third of post-marketing drug withdrawals. Several classifiers for human hepatotoxicity using chemical descriptors with limited prediction accuracies have been published. In this study, we developed predictive in silico models based on a set of 156 DILI positive and 136 DILI negative compounds for DILI prediction. First, models based on a chemical descriptor (CDK, Dragon and MOE) and in vitro cell-imaging endpoints [human hepatocyte imaging assay technology (HIAT) descriptors] were built using random forest (RF) and five-fold cross-validation procedure. Then three hybrid models were built using HIAT and a single type of chemical descriptors. Generally, the models based only on chemical descriptors were poor, with a correct classification rate (CCR) around 0.60 when the default threshold value (i.e. threshold = 0.50) was used. The hybrid models afforded a CCR of 0.73 with a specificity of 0.74 and a better true positive rate (sensitivity of 0.71), which is crucial in drug toxicity screening for the purpose of patient safety. The benefit of hybrid models was even more drastic when stricter classification thresholds were employed (e.g. CCR would be 0.83 when double thresholds (non-toxic < 0.40 and toxic > 0.60) were used for the hybrid model). We have developed rigorously validated hybrid models which can be used in virtual screening of lead compounds with potential hepatotoxicity. Our study also showed a chemical structure and in vitro biological data can be complementary in enhancing the prediction accuracy of human hepatotoxicity and can afford rational mechanistic interpretation.

Comparison of in silico models for prediction of mutagenicity

Using a dataset with more than 6000 compounds, the performance of eight quantitative structure activity relationships (QSAR) models was evaluated: ACD/Tox Suite, Absorption, Distribution, Metabolism, Elimination, and Toxicity of chemical substances (ADMET) predictor, Derek, Toxicity Estimation Software Tool (T.E.S.T.), TOxicity Prediction by Komputer Assisted Technology (TOPKAT), Toxtree, CEASAR, and SARpy (SAR in python). In general, the results showed a high level of performance. To have a realistic estimate of the predictive ability, the results for chemicals inside and outside the training set for each model were considered. The effect of applicability domain tools (when available) on the prediction accuracy was also evaluated. The predictive tools included QSAR models, knowledge-based systems, and a combination of both methods. Models based on statistical QSAR methods gave better results.