Mechanisms of Nitrosamine Mutagenicity and Their Relationship to Rodent Carcinogenic Potency.

A thorough literature review was undertaken to understand how the pathways of N-nitrosamine transformation relate to mutagenic potential and carcinogenic potency in rodents. Empirical and computational evidence indicates that a common radical intermediate is created by CYP-mediated hydrogen abstraction at the α-carbon; it is responsible for both activation, leading to the formation of DNA-reactive diazonium species, and deactivation by denitrosation. There are competing sites of CYP metabolism (e.g., ß-carbon), and other reactive species can form following initial bioactivation, although these alternative pathways tend to decrease rather than enhance carcinogenic potency. The activation pathway, oxidative dealkylation, is a common reaction in drug metabolism and evidence indicates that the carbonyl byproduct, e.g., formaldehyde, does not contribute to the toxic properties of N-nitrosamines. Nitric oxide (NO), a side product of denitrosation, can similarly be discounted as an enhancer of N-nitrosamine toxicity based on carcinogenicity data for substances that act as NO-donors. However, not all N-nitrosamines are potent rodent carcinogens. In a significant number of cases, there is a potency overlap with non-N-nitrosamine carcinogens that are not in the Cohort of Concern (CoC; high-potency rodent carcinogens comprising aflatoxin-like-, N-nitroso-, and alkyl-azoxy compounds), while other N-nitrosamines are devoid of carcinogenic potential. In this context, mutagenicity is a useful surrogate for carcinogenicity, as proposed in the ICH M7 (R2) (2023) guidance. Thus, in the safety assessment and control of N-nitrosamines in medicines, it is important to understand those complementary attributes of mechanisms of mutagenicity and structure-activity relationships that translate to elevated potency versus those which are associated with a reduction in, or absence of, carcinogenic potency.

Deriving acceptable limits for non-mutagenic impurities in medicinal products - Durational adjustments.

ICH Q3A/B guidelines are not intended for application during the clinical research phase of development and durationally adjusted qualification thresholds are not included. A central tenet of ICH Q3A is that lifetime exposure to 1 mg/day of an unqualified non-mutagenic impurity (NMI) is not a safety concern. An analysis of in vivo toxicology data from 4878 unique chemicals with established NO(A)ELs was conducted to determine whether durationally adjusted qualification limits can be supported. Although not recommended in ICH Q3A/B, a conservative approach was taken by using allometric scaling in the analysis. Following allometric scaling of the 5th percentile of the distribution of NO(A)ELs from available chronic toxicology studies, it was reconfirmed that there is a safety basis for the 1 mg/day qualification threshold in ICH Q3A. Additionally, allometric scaling of the 5th percentile of the distribution of NO(A)ELs from sub-acute and sub-chronic toxicology studies could support acceptable limits of 20 and 5 mg/day for an unqualified NMI for dosing durations of less than or greater than one month, respectively. This analysis supports durationally adjusted NMI qualification thresholds for pharmaceuticals that protect patient safety and contribute to 3Rs efforts for qualifying impurities using new approach methods.

Retrospective analysis of the potential use of virtual control groups in preclinical toxicity assessment using the eTOX database.

Virtual Control Groups (VCGs) based on Historical Control Data (HCD) in preclinical toxicity testing have the potential to reduce animal usage. As a case study we retrospectively analyzed the impact of replacing Concurrent Control Groups (CCGs) with VCGs on the treatment-relatedness of 28 selected histopathological findings reported in either rat or dog in the eTOX database. We developed a novel methodology whereby statistical predictions of treatment-relatedness using either CCGs or VCGs of varying covariate similarity to CCGs were compared to designations from original toxicologist reports; and changes in agreement were used to quantify changes in study outcomes. Generally, the best agreement was achieved when CCGs were replaced with VCGs with the highest level of similarity; the same species, strain, sex, administration route, and vehicle. For example, balanced accuracies for rat findings were 0.704 (predictions based on CCGs) vs. 0.702 (predictions based on VCGs). Moreover, we identified covariates which resulted in poorer identification of treatment-relatedness. This was related to an increasing incidence rate divergence in HCD relative to CCGs. Future databases which collect data at the individual animal level including study details such as animal age and testing facility are required to build adequate VCGs to accurately identify treatment-related effects.

Statistical analysis of preclinical inter-species concordance of histopathological findings in the eTOX database.

Preclinical inter-species concordance can increase the predictivity of observations to the clinic, potentially reducing drug attrition caused by unforeseen adverse events. We quantified inter-species concordance of histopathological findings and target organ toxicities across four preclinical species in the eTOX database using likelihood ratios (LRs). This was done whilst only comparing findings between studies with similar compound exposure (Δ|Cmax| ≤ 1 log-unit), repeat-dosing duration, and animals of the same sex. We discovered 24 previously unreported significant inter-species associations between histopathological findings encoded by the HPATH ontology. More associations with strong positive concordance (33% LR+ > 10) relative to strong negative concordance (12.5% LR- < 0.1) were identified. Of the top 10 most positively concordant associations, 60% were computed between different histopathological findings indicating potential differences in inter-species pathogenesis. We also observed low inter-species target organ toxicity concordance. For example, liver toxicity concordance in short-term studies between female rats and dogs observed an average LR+ of 1.84, and an average LR- of 0.73. This was corroborated by similarly low concordance between rodents and non-rodents for 75 candidate drugs in AstraZeneca. This work provides new statistically significant associations between preclinical species, but finds that concordance is rare, particularly between the absence of findings.

Prediction of In Vivo Pharmacokinetic Parameters and Time-Exposure Curves in Rats Using Machine Learning from the Chemical Structure.

Animal pharmacokinetic (PK) data as well as human and animal in vitro systems are utilized in drug discovery to define the rate and route of drug elimination. Accurate prediction and mechanistic understanding of drug clearance and disposition in animals provide a degree of confidence for extrapolation to humans. In addition, prediction of in vivo properties can be used to improve design during drug discovery, help select compounds with better properties, and reduce the number of in vivo experiments. In this study, we generated machine learning models able to predict rat in vivo PK parameters and concentration-time PK profiles based on the molecular chemical structure and either measured or predicted in vitro parameters. The models were trained on internal in vivo rat PK data for over 3000 diverse compounds from multiple projects and therapeutic areas, and the predicted endpoints include clearance and oral bioavailability. We compared the performance of various traditional machine learning algorithms and deep learning approaches, including graph convolutional neural networks. The best models for PK parameters achieved R2 = 0.63 [root mean squared error (RMSE) = 0.26] for clearance and R2 = 0.55 (RMSE = 0.46) for bioavailability. The models provide a fast and cost-efficient way to guide the design of molecules with optimal PK profiles, to enable the prediction of virtual compounds at the point of design, and to drive prioritization of compounds for in vivo assays.

New Associations between Drug-Induced Adverse Events in Animal Models and Humans Reveal Novel Candidate Safety Targets.

To improve our ability to extrapolate preclinical toxicity to humans, there is a need to understand and quantify the concordance of adverse events (AEs) between animal models and clinical studies. In the present work, we discovered 3011 statistically significant associations between preclinical and clinical AEs caused by drugs reported in the PharmaPendium database of which 2952 were new associations between toxicities encoded by different Medical Dictionary for Regulatory Activities terms across species. To find plausible and testable candidate off-target drug activities for the derived associations, we investigated the genetic overlap between the genes linked to both a preclinical and a clinical AE and the protein targets found to interact with one or more drugs causing both AEs. We discuss three associations from the analysis in more detail for which novel candidate off-target drug activities could be identified, namely, the association of preclinical mutagenicity readouts with clinical teratospermia and ovarian failure, the association of preclinical reflexes abnormal with clinical poor-quality sleep, and the association of preclinical psychomotor hyperactivity with clinical drug withdrawal syndrome. Our analysis successfully identified a total of 77% of known safety targets currently tested in in vitro screening panels plus an additional 431 genes which were proposed for investigation as future safety targets for different clinical toxicities. This work provides new translational toxicity relationships beyond AE term-matching, the results of which can be used for risk profiling of future new chemical entities for clinical studies and for the development of future in vitro safety panels.

Machine Learning Models for Human In Vivo Pharmacokinetic Parameters with In-House Validation.

Prior to clinical development, a comprehensive pharmacokinetic characterization of a novel drug is required to understand its exposure at the site of action and elimination. Accordingly, in vitro assays and animal pharmacokinetic studies are regularly employed to predict drug exposure in humans, which is often costly and time-consuming. For this reason, the prediction of human pharmacokinetics at the point of design would be of high value for drug discovery. Therefore, we have established a comprehensive data curation protocol that enables machine learning evaluation of 12 human in vivo pharmacokinetic parameters using only chemical structure information and available doses for 1001 unique compounds. These machine learning models were thoroughly investigated and validated using both an independent hold-out test set and AstraZeneca clinical data. In addition, the availability of preclinical predictions for a subset of internal clinical candidates allowed us to compare our in silico approach with state-of-the-art pharmacokinetic predictions. Based on this evaluation, three fit-for-purpose models for AUC PO (Rtest2 = 0.63; RMSEtest = 0.76), Cmax PO (Rtest2 = 0.68; RMSEtest = 0.62), and Vdss IV (Rtest2 = 0.47; RMSEtest = 0.50) were identified. Based on the findings, our machine learning models have considerable potential for practical applications in drug discovery, such as influencing decision-making in drug discovery projects and progression of drug candidates toward the clinic.

Information-Derived Mechanistic Hypotheses for Structural Cardiotoxicity.

Adverse events resulting from drug therapy can be a cause of drug withdrawal, reduced and or restricted clinical use, as well as a major economic burden for society. To increase the safety of new drugs, there is a need to better understand the mechanisms causing the adverse events. One way to derive new mechanistic hypotheses is by linking data on drug adverse events with the drugs' biological targets. In this study, we have used data mining techniques and mutual information statistical approaches to find associations between reported adverse events collected from the FDA Adverse Event Reporting System and assay outcomes from ToxCast, with the aim to generate mechanistic hypotheses related to structural cardiotoxicity (morphological damage to cardiomyocytes and/or loss of viability). Our workflow identified 22 adverse event-assay outcome associations. From these associations, 10 implicated targets could be substantiated with evidence from previous studies reported in the literature. For two of the identified targets, we also describe a more detailed mechanism, forming putative adverse outcome pathways associated with structural cardiotoxicity. Our study also highlights the difficulties deriving these type of associations from the very limited amount of data available.

Current nonclinical testing paradigms in support of safe clinical trials: An IQ Consortium DruSafe perspective.

The transition from nonclinical to First-in-Human (FIH) testing is one of the most challenging steps in drug development. In response to serious outcomes in a recent Phase 1 trial (sponsored by Bial), IQ Consortium/DruSafe member companies reviewed their nonclinical approach to progress small molecules safely to FIH trials. As a common practice, safety evaluation begins with target selection and continues through iterative in silico and in vitro screening to identify molecules with increased probability of acceptable in vivo safety profiles. High attrition routinely occurs during this phase. In vivo exploratory and pivotal FIH-enabling toxicity studies are then conducted to identify molecules with a favorable benefit-risk profile for humans. The recent serious incident has reemphasized the importance of nonclinical testing plans that are customized to the target, the molecule, and the intended clinical plan. Despite the challenges and inherent risks of transitioning from nonclinical to clinical testing, Phase 1 studies have a remarkably good safety record. Given the rapid scientific evolution of safety evaluation, testing paradigms and regulatory guidance must evolve with emerging science. The authors posit that the practices described herein, together with science-based risk assessment and management, support safe FIH trials while advancing development of important new medicines.

It's difficult, but important, to make negative predictions.

At the confluence of predictive and regulatory toxicologies, negative predictions may be the thin green line that prevents populations from being exposed to harm. Here, two novel approaches to making confident and robust negative in silico predictions for mutagenicity (as defined by the Ames test) have been evaluated. Analyses of 12 data sets containing >13,000 compounds, showed that negative predictivity is high (∼90%) for the best approach and features that either reduce the accuracy or certainty of negative predictions are identified as misclassified or unclassified respectively. However, negative predictivity remains high (and in excess of the prevalence of non-mutagens) even in the presence of these features, indicating that they are not flags for mutagenicity.

Evaluation of a statistics-based Ames mutagenicity QSAR model and interpretation of the results obtained.

The relative wealth of bacterial mutagenicity data available in the public literature means that in silico quantitative/qualitative structure activity relationship (QSAR) systems can readily be built for this endpoint. A good means of evaluating the performance of such systems is to use private unpublished data sets, which generally represent a more distinct chemical space than publicly available test sets and, as a result, provide a greater challenge to the model. However, raw performance metrics should not be the only factor considered when judging this type of software since expert interpretation of the results obtained may allow for further improvements in predictivity. Enough information should be provided by a QSAR to allow the user to make general, scientifically-based arguments in order to assess and overrule predictions when necessary. With all this in mind, we sought to validate the performance of the statistics-based in vitro bacterial mutagenicity prediction system Sarah Nexus (version 1.1) against private test data sets supplied by nine different pharmaceutical companies. The results of these evaluations were then analysed in order to identify findings presented by the model which would be useful for the user to take into consideration when interpreting the results and making their final decision about the mutagenic potential of a given compound.

A practical application of two in silico systems for identification of potentially mutagenic impurities.

The International Conference on Harmonization (ICH) M7 guidance for the assessment and control of DNA reactive impurities in pharmaceutical products includes the use of in silico prediction systems as part of the hazard identification and risk assessment strategy. This is the first internationally agreed guidance document to include the use of these types of approaches. The guideline requires the use of two complementary approaches, an expert rule-based method and a statistical algorithm. In addition, the guidance states that the output from these computer-based assessments can be reviewed using expert knowledge to provide additional support or resolve conflicting predictions. This approach is designed to maximize the sensitivity for correctly identifying DNA reactive compounds while providing a framework to reduce the number of compounds that need to be synthesized, purified and subsequently tested in an Ames assay. Using a data set of 801 chemicals and pharmaceutical intermediates, we have examined the relative predictive performances of some popular commercial in silico systems that are in common use across the pharmaceutical industry. The overall accuracy of each of these systems was fairly comparable ranging from 68% to 73%; however, the sensitivity of each system (i.e. how many Ames positive compounds are correctly identified) varied much more dramatically from 48% to 68%. We have explored how these systems can be combined under the ICH M7 guidance to enhance the detection of DNA reactive molecules. Finally, using four smaller sets of molecules, we have explored the value of expert knowledge in the review process, especially in cases where the two systems disagreed on their predictions, and the need for care when evaluating the predictions for large data sets.

(Q)SAR: A Tool for the Toxicologist.

A continuing education (CE) course at the 2014 American College of Toxicology annual meeting covered the topic of (Quantitative) Structure-Activity Relationships [(Q)SAR]. The (Q)SAR methodologies use predictive computer modeling based on predefined rules to describe the relationship between chemical structure and a chemical's associated biological activity or statistical tools to find correlations between biologic activity and the molecular structure or properties of a compound. The (Q)SAR has applications in risk assessment, drug discovery, and regulatory decision making. Pressure within industry to reduce the cost of drug development and societal pressure for government regulatory agencies to produce more accurate and timely risk assessment of drugs and chemicals have necessitated the use of (Q)SAR. Producing a high-quality (Q)SAR model depends on many factors including the choice of statistical methods and descriptors, but first and foremost the quality of the data input into the model. Understanding how a (Q)SAR model is developed and applied is critical to the successful use of such a tool. The CE session covered the basic principles of (Q)SAR, practical applications of these computational models in toxicology, how regulatory agencies use and interpret (Q)SAR models, and potential pitfalls of using them.

Analysis of Pfizer compounds in EPA's ToxCast chemicals-assay space.

The U.S. Environmental Protection Agency (EPA) launched the ToxCast program in 2007 with the goal of evaluating high-throughput in vitro assays to prioritize chemicals that need toxicity testing. Their goal was to develop predictive bioactivity signatures for toxic compounds using a set of in vitro assays and/or in silico properties. In 2009, Pfizer joined the ToxCast initiative by contributing 52 compounds with preclinical and clinical data for profiling across the multiple assay platforms available. Here, we describe the initial analysis of the Pfizer subset of compounds within the ToxCast chemical (n = 1814) and in vitro assay (n = 486) space. An analysis of the hit rate of Pfizer compounds in the ToxCast assay panel allowed us to focus our mining of assays potentially most relevant to the attrition of our compounds. We compared the bioactivity profile of Pfizer compounds to other compounds in the ToxCast chemical space to gain insights into common toxicity pathways. Additionally, we explored the similarity in the chemical and biological spaces between drug-like compounds and environmental chemicals in ToxCast and compared the in vivo profiles of a subset of failed pharmaceuticals having high similarity in both spaces. We found differences in the chemical and biological spaces of pharmaceuticals compared to environmental chemicals, which may question the applicability of bioactivity signatures developed exclusively based on the latter to drug-like compounds if used without prior validation with the ToxCast Phase-II chemicals. Finally, our analysis has allowed us to identify novel interactions for our compounds in particular with multiple nuclear receptors that were previously not known. This insight may help us to identify potential liabilities with future novel compounds.

Chemotypes sensitivity and predictivity of in vivo outcomes for cytotoxic assays in THLE and HepG2 cell lines.

In the present study, we demonstrate the utility of in vitro ATP depletion assays in both THLE and HepG2 cells for predicting the toxicological outcome in Exploratory Toxicology Studies across 446 Pfizer proprietary compounds. Our results suggest a higher likelihood of selecting suitable compounds for in vivo safety studies by using cytotoxicity assays in multiple cell-lines over a single cell line. In addition, we demonstrate that different cell-lines have different sensitivities to compounds depending on their ionization state, that is, acid, base or neutral. HepG2 cells are more sensitive for basic compounds, whereas THLE cells have a relatively higher sensitivity for the acidic and neutral compounds. These in vitro cytotoxicity assays when combined with physicochemical properties (cLogP >3 and topological polar surface area (TPSA) <75Å(2)), are the most effective means to prioritize compounds having a lower probability of causing adverse events in vivo.

A Decade in a Systematic Review: The Evolution and Impact of Cell Painting.

High-content image-based assays have fueled significant discoveries in the life sciences in the past decade (2013-2023), including novel insights into disease etiology, mechanism of action, new therapeutics, and toxicology predictions. Here, we systematically review the substantial methodological advancements and applications of Cell Painting. Advancements include improvements in the Cell Painting protocol, assay adaptations for different types of perturbations and applications, and improved methodologies for feature extraction, quality control, and batch effect correction. Moreover, machine learning methods recently surpassed classical approaches in their ability to extract biologically useful information from Cell Painting images. Cell Painting data have been used alone or in combination with other -omics data to decipher the mechanism of action of a compound, its toxicity profile, and many other biological effects. Overall, key methodological advances have expanded Cell Painting's ability to capture cellular responses to various perturbations. Future advances will likely lie in advancing computational and experimental techniques, developing new publicly available datasets, and integrating them with other high-content data types.

A Decade in a Systematic Review: The Evolution and Impact of Cell Painting.

High-content image-based assays have fueled significant discoveries in the life sciences in the past decade (2013-2023), including novel insights into disease etiology, mechanism of action, new therapeutics, and toxicology predictions. Here, we systematically review the substantial methodological advancements and applications of Cell Painting. Advancements include improvements in the Cell Painting protocol, assay adaptations for different types of perturbations and applications, and improved methodologies for feature extraction, quality control, and batch effect correction. Moreover, machine learning methods recently surpassed classical approaches in their ability to extract biologically useful information from Cell Painting images. Cell Painting data have been used alone or in combination with other - omics data to decipher the mechanism of action of a compound, its toxicity profile, and many other biological effects. Overall, key methodological advances have expanded Cell Painting's ability to capture cellular responses to various perturbations. Future advances will likely lie in advancing computational and experimental techniques, developing new publicly available datasets, and integrating them with other high-content data types.

Use of in silico systems and expert knowledge for structure-based assessment of potentially mutagenic impurities.

Genotoxicity hazard identification is part of the impurity qualification process for drug substances and products, the first step of which being the prediction of their potential DNA reactivity using in silico (quantitative) structure-activity relationship (Q)SAR models/systems. This white paper provides information relevant to the development of the draft harmonized tripartite guideline ICH M7 on potentially DNA-reactive/mutagenic impurities in pharmaceuticals and their application in practice. It explains relevant (Q)SAR methodologies as well as the added value of expert knowledge. Moreover, the predictive value of the different methodologies analyzed in two surveys conveyed in the US and European pharmaceutical industry is compared: most pharmaceutical companies used a rule-based expert system as their primary methodology, yielding negative predictivity values of ⩾78% in all participating companies. A further increase (>90%) was often achieved by an additional expert review and/or a second QSAR methodology. Also in the latter case, an expert review was mandatory, especially when conflicting results were obtained. Based on the available data, we concluded that a rule-based expert system complemented by either expert knowledge or a second (Q)SAR model is appropriate. A maximal transparency of the assessment process (e.g. methods, results, arguments of weight-of-evidence approach) achieved by e.g. data sharing initiatives and the use of standards for reporting will enable regulators to fully understand the results of the analysis. Overall, the procedures presented here for structure-based assessment are considered appropriate for regulatory submissions in the scope of ICH M7.

In silico methods combined with expert knowledge rule out mutagenic potential of pharmaceutical impurities: an industry survey.

With the increasing emphasis on identification and low level control of potentially genotoxic impurities (GTIs), there has been increased use of structure-based assessments including application of computerized models. To date many publications have focused on the ability of computational models, either individually or in combination, to accurately predict the mutagenic effects of a chemical in the Ames assay. Typically, these investigations take large numbers of compounds and use in silico tools to predict their activity with no human interpretation being made. However, this does not reflect how these assessments are conducted in practice across the pharmaceutical industry. Current guidelines indicate that a structural assessment is sufficient to conclude that an impurity is non-mutagenic. To assess how confident we can be in identifying non-mutagenic structures, eight companies were surveyed for their success rate. The Negative Predictive Value (NPV) of the in silico approaches was 94%. When human interpretation of in silico model predictions was conducted, the NPV increased substantially to 99%. The survey illustrates the importance of expert interpretation of in silico predictions. The survey also suggests the use of multiple computational models is not a significant factor in the success of these approaches with respect to NPV.