The database makes the poison: How the selection of datasets in QSAR models impacts toxicant prediction of higher tier endpoints.

As the United States and the European Union continue their steady march towards the acceptance of new approach methodologies (NAMs), we need to ensure that the available tools are fit for purpose. Critics will be well-positioned to caution against NAMs acceptance and adoption if the tools turn out to be inadequate. In this paper, we focus on Quantitative Structure Activity-Relationship (QSAR) models and highlight how the training database affects quality and performance of these models. Our analysis goes to the point of asking, "are the endpoints extracted from the experimental studies in the database trustworthy, or are they false negatives/positives themselves?" We also discuss the impacts of chemistry on QSAR models, including issues with 2-D structure analyses when dealing with isomers, metabolism, and toxicokinetics. We close our analysis with a discussion of challenges associated with translational toxicology, specifically the lack of adverse outcome pathways/adverse outcome pathway networks (AOPs/AOPNs) for many higher tier endpoints. We recognize that it takes a collaborate effort to build better and higher quality QSAR models especially for higher tier toxicological endpoints. Hence, it is critical to bring toxicologists, statisticians, and machine learning specialists together to discuss and solve these challenges to get relevant predictions.

A new approach methodology using kinetically-derived maximum dose levels in risk assessment - A case study with afidopyropen.

We present a case study for afidopyropen (AF; insecticide) to characterize chronic dietary human health risk using a Risk 21-based approach. Our objective is to use a well-tested pesticidal active ingredient (AF) to show how a new approach methodology (NAM), using the kinetically-derived maximum dose (KMD) and with far less animal testing, can reliably identify a health-protective point of departure (PoD) for chronic dietary human health risk assessments (HHRA). Chronic dietary HHRA involves evaluation of both hazard and exposure information to characterize risk. Although both are important, emphasis has been placed on a checklist of required toxicological studies for hazard characterization, with human exposure information only considered after evaluation of hazard data. Most required studies are not used to define the human endpoint for HHRA. The information presented demonstrates a NAM that uses the KMD determined by saturation of a metabolic pathway, which can be used as an alternative POD. In these cases, the full toxicological database may not need to be generated. Demonstration that the compound is not genotoxic and that the KMD is protective of adverse effects in 90-day oral rat and reproductive/developmental studies is sufficient to support the use of the KMD as an alternative POD.

Computational prediction of the metabolites of agrochemicals formed in rats.

Today, computational tools for the prediction of the metabolite structures of xenobiotics are widely available and employed in small-molecule research. Reflecting the availability of measured data, these in silico tools are trained and validated primarily on drug metabolism data. In this work, we assessed the capacity of five leading metabolite structure predictors to represent the metabolism of agrochemicals observed in rats. More specifically, we tested the ability of SyGMa, GLORY, GLORYx, BioTransformer 3.0, and MetaTrans to correctly predict and rank the experimentally observed metabolites of a set of 85 parent compounds. We found that the models were able to recover about one to two-thirds of the experimentally observed first-generation, second-generation and third-generation metabolites, confirming their value in applications such as metabolite identification. However, precision was low for all investigated tools and did not exceed approximately 18 % for the pool of first-generation metabolites and 2 % for the pool of compounds representing the first three generations of metabolites. The variance in prediction success rates was high across the individual metabolic maps, meaning that outcomes depend strongly on the specific compound under investigation. We also found that the predictions for individual parent compounds differed strongly between the tools, particularly between those built on orthogonal technologies (e.g., rule-based and end-to-end machine learning approaches). This renders ensemble model strategies promising for improving success rates. Overall, the results of this benchmark study show that there is still considerable room for the improvement of metabolite structure predictors left. Our discussion points out several avenues to progress. The bottleneck in method development certainly has been, and will remain, for the foreseeable future, the limited quantity and quality of available measured data on small-molecule metabolism.

Making in silico predictive models for toxicology FAIR.

In silico predictive models for toxicology include quantitative structure-activity relationship (QSAR) and physiologically based kinetic (PBK) approaches to predict physico-chemical and ADME properties, toxicological effects and internal exposure. Such models are used to fill data gaps as part of chemical risk assessment. There is a growing need to ensure in silico predictive models for toxicology are available for use and that they are reproducible. This paper describes how the FAIR (Findable, Accessible, Interoperable, Reusable) principles, developed for data sharing, have been applied to in silico predictive models. In particular, this investigation has focussed on how the FAIR principles could be applied to improved regulatory acceptance of predictions from such models. Eighteen principles have been developed that cover all aspects of FAIR. It is intended that FAIRification of in silico predictive models for toxicology will increase their use and acceptance.

Understanding and overcoming the technical challenges in using in silico predictions in regulatory decisions of complex toxicological endpoints - A pesticide perspective for regulatory toxicologists with a focus on machine learning models.

There are many challenges that must be overcome before in silico toxicity predictions are ripe for regulatory decision-making. Today, mandates in the United States of America and the European Union to avoid animal usage in toxicity testing is driving the need to consider alternative technologies, including Quantitative Structure Activity Relationship (QSAR) models, and read across approaches. However, when adopting new methods, it is critical that both new approach developers as well as regulatory users understand the strengths and challenges with these new approaches. In this paper, we identify potential sources of bias in machine learning methods specific to toxicity predictions, that may impact the overall performance of in silico models. We also discuss ways to mitigate these biases. Based on our experiences, the most prevalent sources of bias include class imbalance (differing numbers of "toxic" vs "nontoxic" compounds), limited numbers of chemicals within a particular chemistry, and biases within the studies that make up the database used for model building, as well as model evaluation biases. While this is already complex for repeated dose toxicity, in reproduction and developmental toxicity a further level of complexity is introduced by the need to evaluate effects on individual animal and litter basis (e.g., a hierarchal structure). We also discuss key considerations developers and regulators need to make when they use machine learning models to predict chemical safety. Our objective is for our paper to serve as a desk reference for model developers and regulators as they evaluate machine learning models and as they make decisions using these models.

Review of the state of science and evaluation of currently available in silico prediction models for reproductive and developmental toxicity: A case study on pesticides.

BACKGROUND: In silico methods for toxicity prediction have increased significantly in recent years due to the 3Rs principle. This also applies to predicting reproductive toxicology, which is one of the most critical factors in pesticide approval. The widely used quantitative structure-activity relationship (QSAR) models use experimental toxicity data to create a model that relates experimentally observed toxicity to molecular structures to predict toxicity. Aim of the study was to evaluate the available prediction models for developmental and reproductive toxicity regarding their strengths and weaknesses in a pesticide database. METHODS: The reproductive toxicity of 315 pesticides, which have a GHS classification by ECHA, was compared with the prediction of different in silico models: VEGA, OECD (Q)SAR Toolbox, Leadscope Model Applier, and CASE Ultra by MultiCASE. RESULTS: In all models, a large proportion (up to 77%) of all pesticides were outside the chemical space of the model. Analysis of the prediction of remaining pesticides revealed a balanced accuracy of the models between 0.48 and 0.66. CONCLUSION: Overall, predictions were only meaningful in rare cases and therefore always require evaluation by an expert. The critical factors were the underlying data and determination of molecular similarity, which offer great potential for improvement.

Evaluating Confidence in Toxicity Assessments Based on Experimental Data and In Silico Predictions.

Understanding the reliability and relevance of a toxicological assessment is important for gauging the overall confidence and communicating the degree of uncertainty related to it. The process involved in assessing reliability and relevance is well defined for experimental data. Similar criteria need to be established for in silico predictions, as they become increasingly more important to fill data gaps and need to be reasonably integrated as additional lines of evidence. Thus, in silico assessments could be communicated with greater confidence and in a more harmonized manner. The current work expands on previous definitions of reliability, relevance, and confidence and establishes a conceptional framework to apply those to in silico data. The approach is used in two case studies: 1) phthalic anhydride, where experimental data are readily available and 2) 4-hydroxy-3-propoxybenzaldehyde, a data poor case which relies predominantly on in silico methods, showing that reliability, relevance, and confidence of in silico assessments can be effectively communicated within Integrated approaches to testing and assessment (IATA).

Ontogeny of renal, hepatic, and placental expression of ATP-binding cassette and solute carrier transporters in the rat and the rabbit.

Species differences in developmental toxicity can be due to varying expression of xenobiotic transporters. Hence, knowledge on the ontogeny of these transporters, especially in human, rat and rabbit, is pivotal. Two superfamilies of transporters, the ATP-binding cassette (ABC) and the solute carrier (SLC) transporters, are well known for their role in the absorption, distribution and/or elimination of xenobiotics and endogenous substances. The aim of this study was to compare the expression levels of these xenobiotic transporters in liver, kidney and placenta of man, Wistar rat and New Zealand White rabbit during pre- and postnatal development. For this purpose, qPCR experiments were performed for rat and rabbit tissues and the gene expression profiles were compared with literature data from man, rat and rabbit. Data analysis showed large differences in transporter expression in development and between species. These results can be used to better understand developmental toxicity findings in non-clinical species and their relevance for man.

Afidopyropen: Challenges and impact of a toxicokinetic study designed to identify a point of inflection from dose-proportionality.

Afidopyropen is an insecticide that acts as a transient receptor potential vanilloid subtype (TRPV) channel modulator in chordotonal organs of target insects and has been assessed for a wide range of toxicity endpoints including chronic toxicity and carcinogenicity in rats and mice. The current study evaluates the toxicokinetic properties of afidopyropen and its plasma metabolites in rats at dose levels where the pharmacokinetics (PK) are linear and nonlinear in an attempt to identify a point of inflection. Based on the results of this study and depending on the analysis method used, the kinetically derived maximum dose (KMD) is estimated to be between 2.5 and 12.5 mg/kg bw/d with linearity observed at doses below 2.5 mg/kg bw/d. A defined point of inflection could not be determined. These data demonstrate that consideration of PK is critical for improving the dose-selection in toxicity studies as well as to enhance human relevance of the interpretation of animal toxicity studies. The study also demonstrates the technical difficulty in obtaining a defined point of inflection from in vivo PK data.

In Silico Approaches In Carcinogenicity Hazard Assessment: Current Status and Future Needs.

Historically, identifying carcinogens has relied primarily on tumor studies in rodents, which require enormous resources in both money and time. In silico models have been developed for predicting rodent carcinogens but have not yet found general regulatory acceptance, in part due to the lack of a generally accepted protocol for performing such an assessment as well as limitations in predictive performance and scope. There remains a need for additional, improved in silico carcinogenicity models, especially ones that are more human-relevant, for use in research and regulatory decision-making. As part of an international effort to develop in silico toxicological protocols, a consortium of toxicologists, computational scientists, and regulatory scientists across several industries and governmental agencies evaluated the extent to which in silico models exist for each of the recently defined 10 key characteristics (KCs) of carcinogens. This position paper summarizes the current status of in silico tools for the assessment of each KC and identifies the data gaps that need to be addressed before a comprehensive in silico carcinogenicity protocol can be developed for regulatory use.

Use of toxicokinetic data for afidopyropen to determine the dose levels in developmental toxicity studies.

Afidopyropen is an insecticide that acts as a TRPV channel modulator in chordotonal organs of target insects and has been assessed for a wide range of toxicity endpoints including developmental toxicity in rats and rabbits. The GLP developmental toxicity study in rabbits did not produce evidence of maternal or fetal toxicity at the highest dose tested (32 mg/kg/day) but pharmacokinetics (PK) in pregnant rabbits in this study exhibited onset of PK nonlinearity from 5 mg/kg/day on, as measured by plasma Cmax and AUC. The NOAEL (32 mg/kg/day) is 9000X higher than maximum expected human dietary exposures to afidopyropen; the dose range where nonlinear PK were observed (5-15 mg/kg/day) is 1400-4200X higher. As nonlinearity occurred between 5 and 15 mg/kg/day, 32 mg/kg/day is concluded to be a sufficiently high dose (kinetically derived maximum dose) for a prenatal developmental toxicity study. As recognized by regulatory dose-selection guidance, onset of saturated PK is evidence of excessive biological stress to test animals rendering any effects at such doses of questionable relevance for human risk assessment. These data demonstrate that consideration of PK is critical for improving the dose-selection in developmental toxicity studies to enhance human relevance of animal toxicity studies.

Genetic toxicology in silico protocol.

In silico toxicology (IST) approaches to rapidly assess chemical hazard, and usage of such methods is increasing in all applications but especially for regulatory submissions, such as for assessing chemicals under REACH as well as the ICH M7 guideline for drug impurities. There are a number of obstacles to performing an IST assessment, including uncertainty in how such an assessment and associated expert review should be performed or what is fit for purpose, as well as a lack of confidence that the results will be accepted by colleagues, collaborators and regulatory authorities. To address this, a project to develop a series of IST protocols for different hazard endpoints has been initiated and this paper describes the genetic toxicity in silico (GIST) protocol. The protocol outlines a hazard assessment framework including key effects/mechanisms and their relationships to endpoints such as gene mutation and clastogenicity. IST models and data are reviewed that support the assessment of these effects/mechanisms along with defined approaches for combining the information and evaluating the confidence in the assessment. This protocol has been developed through a consortium of toxicologists, computational scientists, and regulatory scientists across several industries to support the implementation and acceptance of in silico approaches.

Uterine adenocarcinoma in the rat induced by afidopyropen. An analysis of the lesion's induction, progression and its relevance to humans.

Afidopyropen is a novel insecticide that acts as a TRPV channel modulator in chordotonal organs of target insects. In two carcinogenicity studies with Fischer rats, an increased incidence of uterine adenocarcinomas was observed at 1000 and 3000 ppm. This finding prompted an investigation of the mechanism of the tumor formation as well as the relevance of this mechanism to humans. The mechanistic work took parallel paths: one path investigated the pharmacokinetic properties of the test substance at the doses where the tumors were found; while the second path examined the key mechanistic events that culminated in uterine adenocarcinomas. The results of the investigation indicated that the tumors only occurred at doses where excretion of test substance was saturated - indicating that homeostatic biological and/or physiological processes were overwhelmed. At the doses where these processes were overwhelmed, the test substance acted through a mechanism of dopamine agonism, triggering a cascade key events that resulted in uterine adenocarcinomas. An analysis of these mechanisms observed in rat showed that they are both quantitatively (pharmacokinetic mechanism) and qualitatively (dopamine agonism mechanism) not relevant to humans. Therefore the uterine adenocarcinomas observed in the rat associated with high doses of Afidopyropen are not expected to pose a carcinogenic risk to humans.

The arylhydrocarbon receptor is only marginally involved in the antileukemic effects of its ligand curcumin.

BACKGROUND: Acute myeloid leukaemia (AML) continues to present demanding treatment challenges, as in general the prognosis for long-term survival remains dire for the patients. Natural plant-derived substances with antileukemic properties offer new treatment possibilities or may act as by-stander therapy. Their molecular mechanisms of action are often not entirely clear, limiting theory-directed screening and application strategies. The plant substance curcumin is a known activator of the transcription factor aryl hydrocarbon receptor (AhR), and has well-documented antileukemic effects. The AhR regulates cell processes, including cell cycle and apoptosis. We ask here whether direct AhR-activation by curcumin contributes to its antileukemic/apoptotic potential. MATERIALS AND METHODS: The induction of caspases 3/7, 8, and 9, the breakdown of mitochondrial transmembrane potential, the BCL-2/BAX ratio, and the DNA content of cells were measured as indicators of apoptosis. In addition, the induction of cell cycle inhibitors p21 and p27 were assessed. RESULTS: While triggering of AhR signalling by curcumin in HL-60 cells was confirmed, induction of the above apoptosis parameters was not blocked by two AhR antagonists, alpha-naphtoflavone (alphaNF) and 3'-methoxy-4'nitroflavone (MNF). Only a moderate (20%) AhR-dependent induction of caspases 3/7 was detectable. Interestingly, transcriptional changes induced by curcumin and by anticarcinogenic 1,25-dihydroxy vitamin D3 overlapped by one third. CONCLUSION: We conclude that AhR is only marginally involved in the antileukemic effects of its ligand curcumin.

A toolbox of novel murine house-keeping genes identified by meta-analysis of large scale gene expression profiles.

Quantitative and semi-quantitative analysis of gene transcripts requires normalization to RNA-input and/or invariantly expressed house-keeping genes. Currently, only a limited choice of reference genes exists, such as GAPDH(1), beta-actin, or HPRT, whose transcription levels may be less stable than previously thought. We used the meta-database NC-GED, which we had derived from 1968 published murine expression profiles to identify genes with (i) low inter-tissue expression variability and (ii) great stability over 312 conditions, such as experimental drug treatment, age or differentiation. We identified 276 novel genes with "house-keeping" characteristics, including many genes for ribosomal proteins, and aryl-hydrocarbon receptor-interacting protein. Most genes yielded medium to strong fluorescence intensity on the arrays, a relative measure for their cellular expression. We validated the invariant expression levels of eight of the house-keeper candidates in lymph nodes, thymus, liver, kidney and brain of four different mouse strains. In addition, comparative analysis showed the superiority of multiple over single standardization. Caution against established reference genes is justified. The new panel of reference genes is useful for a flexible selection of reference genes in gene expression studies.

Promoter analysis of TCDD-inducible genes in a thymic epithelial cell line indicates the potential for cell-specific transcription factor crosstalk in the AhR response.

Activation of the aryl hydrocarbon receptor (AhR(1)) by 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) elicits severe immunosuppression accompanied by thymic atrophy. Previous evidence suggests that TCDD targets both thymocytes and thymic epithelial cells. The AhR induces cell-specific changes in gene transcription via binding to the dioxin response element DRE; however, the underlying specificity-mechanisms, in particular with regard to the role of promoter element context, and possible transcription factor crosstalk remain poorly understood. Global gene expression in the cortical thymic epithelial cell line ET at 2, 4, and 6 h following 5 nM TCDD exposure resulted in differential regulation of 201 genes. JASPAR and TRANSFAC mapped the statistically over-represented promoter elements in the regulated genes to specific transcription factor binding sites, suggesting a regulatory role in AhR signaling. Over-represented elements included the xenobiotic response element XRE, NF kappaB-Rel, HRE, PPAR gamma, GR, PAX-4 and estrogen receptor binding sites. Co-treatment experiments with TCDD and CoCl(2), to induce hypoxia, or TCDD and 17-beta-estradiol (E2) indicated crosstalk between AhR and Hif or ER, in agreement with other experimental models. The computational identification of TFBS and the demonstration of interaction confirm their interactions with AhR signaling and suggest that the other over-represented elements may also be important in the immunosuppressive effects elicited by TCDD. In conclusion, we demonstrated the importance of promoter element cooperation in the shaping of a cell-specific AhR response. Our findings regarding the transcriptional changes in cortical epithelial cells are congruent with the well-known thymotoxic TCDD-phenotype, and useful in new hypothesis generation of the role of cortical TECs in TCDD toxicity.

Microarray analysis of the AHR system: tissue-specific flexibility in signal and target genes.

Data mining published microarray experiments require that expression profiles are directly comparable. We performed linear global normalization on the data of 1967 Affymetrix U74av2 microarrays, i.e. the transcriptomes of >100 murine tissues or cell types. The mathematical transformation effectively nullifies inter-experimental or inter-laboratory differences between microarrays. The correctness of expression values was validated by quantitative RT-PCR. Using the database we analyze components of the aryl hydrocarbon receptor (AHR) signaling pathway in various tissues. We identified lineage and differentiation specific variant expression of AHR, ARNT, and HIF1alpha in the T-cell lineage and high expression of CYP1A1 in immature B cells and dendritic cells. Performing co-expression analysis we found unorthodox expression of the AHR in the absence of ARNT, particularly in stem cell populations, and can reject the hypothesis that ARNT2 takes over and is highly expressed when ARNT expression is low or absent. Furthermore the AHR shows no co-expression with any other transcript present on the chip. Analysis of differential gene expression under 308 conditions revealed 53 conditions under which the AHR is regulated, numerous conditions under which an intrinsic AHR action is modified as well as conditions activating the AHR even in the absence of known AHR ligands. Thus meta-analysis of published expression profiles is a powerful tool to gain novel insights into known and unknown systems.

Transcriptional signatures of immune cells in aryl hydrocarbon receptor (AHR)-proficient and AHR-deficient mice.

The ligand-activated aryl hydrocarbon receptor (AHR) is known to modulate many genes in a highly cell-specific manner, either directly or indirectly via secondary effects. In contrast, little is known about the effects of AHR deficiency on gene expression balance. We compared the transcriptome of CD4 T cells from AHR-/- mice and wild-type mice; 390 genes, many of them immunotypic, were deregulated in AHR-deficient CD4 cells. TCDD-induced transcriptome changes correlated with the AHR expression level in immune cells. However, there was little overlap in AHR-dependent transcripts found in T lineage cells or dendritic cells. Our results demonstrate flexible gene accessibility for the AHR in immune cells. The idea of a universal battery of AHR-responsive genes is not tenable.

Detection of a novel population of fetal thymocytes characterized by preferential emigration and a TCRgammadelta+ T cell fate after dioxin exposure.

T cell maturation into TCRalphabeta(+) or TCRgammadelta(+) cells from common immature CD4(-)CD8(-)(DN) precursors occurs in the thymus, and is controlled through ordered regulation of genes. The aryl hydrocarbon receptor (AHR), a latent cytoplasmic transcription factor, affects thymocyte maturation and differentiation at several stages, also including DN cells. We analyzed in murine fetal thymus organ cultures (FTOC) the outcome of AHR-signaling and found a higher frequency of DN TCRgammadelta(+) cells in the presence of the AHR-activating ligand TCDD. We detected a novel population of CD25(int/lo)CD44(hi) cells associated with preferential emigration and a TCRgammadelta(+) T cell fate of thymocytes. Sorted DN TCRgammadelta(+) emigrants could proliferate if IL-2 was available. Moreover, they suppressed the proliferation of co-cultivated, activated CD4(+) T cells. Gene expression profiles of purified DN emigrants from TCDD*FTOC revealed 295 modulated genes, 10% of which are genes of the immune system. For instance, RAG-1, TdT, and Gfi-1 were downregulated, yet genes indicative of mature thymocytes were upregulated. In conclusion, we have detected changes in the differentiation programme of fetal DN thymocytes after ligand-activation of the AHR. In particular, we observed a higher frequency of DN TCRgammadelta(+) cells with high emigration potential, and possible regulatory functions.