A comprehensive model for reproductive and developmental toxicity hazard identification: II. Construction of QSAR models to predict activities of untested chemicals☆
Introduction
This report describes part II of an investigation to develop a model for the identification of human reproductive and developmental toxicity hazards. This investigation was conducted by the Food and Drug Administration (FDA), Center for Drug Evaluation and Research (CDER), Office of Pharmaceutical Science (OPS), Informatics and Computational Safety Analysis Staff (ICSAS), which is an applied regulatory research unit that compiles databases of the results of the clinical and toxicological studies submitted by the pharmaceutical industry and develops quantitative structure–activity relationships (QSARs) based on these results. (http://www.fda.gov/cder/Offices/OPS_IO/default.htm).
The primary goal of this investigation was to develop a weight of evidence (WOE)-based battery of QSARs to estimate the potential reprotox activities of untested chemicals. Furthermore, it was the objective of this investigation to validate and optimize the predictive performance of the QSARs based upon OECD (Q)SAR guidelines and make them available for regulatory and industry applications. The reprotox QSAR battery was constructed for general reprotox classes using clusters of related specific reprotox categories identified in the companion report (Matthews et al., 2007). Rather than developing hundreds of individual QSAR models for every site-specific reprotox endpoint in each mammalian species, we optimized the QSAR models for the identification of trans-species reprotoxicants that are most likely to have potential human risk and thus high regulatory significance. Our working hypothesis is that chemicals that induce trans-species reproductive toxicity findings in female and male mammals, and trans-species developmental toxicity findings in the mammalian fetus are less influenced by genetic variability and are most likely to be reprotoxicants in humans. Conversely, chemicals that induce reprotox in a single mammalian tissue, gender, and/or species are more influenced by genetic variability among mammalian species and are less likely to be reprotoxicants in humans based upon the results of the animal studies.
We elected to use the MC4PC software for this investigation because we have already demonstrated that it is able to accurately predict the potential carcinogenicity (Matthews and Contrera, 1998) and genetic toxicity (Matthews et al., 2006a, Matthews et al., 2006b) of chemicals. We hypothesized that MC4PC might also be able to identify trans-species structural alerts for general reprotox classes. In addition, other laboratories have used earlier versions of this program (MCASE and CASE) to screen chemicals for potential teratogenicity (Ghanooni et al., 1997, Gomez et al., 1999, Cunningham and Rosenkranz, 2001) and to predict teratogenicity of restricted classes of chemicals (Klopman and Ptchelintsev, 1993).
The contents of this report and the companion report are designed to fulfill the five OECD principles to facilitate the consideration of a (Q)SAR model for regulatory purposes (OECD Environment Health and Safety Publications. XX, 2006). This investigation had: (1) defined endpoints; (2) an unambiguous algorithm; (3) a defined domain of applicability; (4) appropriate measures of goodness-of-fit, robustness and predictivity; and (5) a mechanistic interpretation. Because reprotox is associated with numerous chemical mechanisms, we developed more general (Q)SARs based upon the same observed toxic effect associated with non-congeneric classes of chemicals and different toxicity mechanisms in order to achieve the largest possible applicability domain (AD). The QSAR battery developed in this study is designed to provide QSAR estimates for the reproductive toxicity and developmental toxicity endpoints associated with OECD test guidelines and thereby be useful to REACH (van der Jagt et al., 2003; http://www.europa.eu.int/comm/environmental/chemicals/reach.htm; http://ecb.jrc.it/REACH/), the 7th Amendment of the Cosmetics Directive (http://www.eceae.org/english/cosmetics.html), and other regulatory applications.
Section snippets
Data sources and defined endpoints
A detailed description of the sources of the reproductive and developmental toxicity (reprotox) study data used in this investigation is provided in the companion report (Matthews et al., 2007). Briefly, the study findings were subdivided into seven general reprotox classes: (1) male reproductive toxicity, (2) female reproductive toxicity, (3) fetal dysmorphogenesis, (4) fetal functional toxicity, (5) fetal and newborn mammal mortality, (6) fetal growth, and (7) newborn behavioral toxicity. The
Optimization of QSAR models for general reprotox classes
QSAR validation experiments were performed to optimize detection of the general reproductive and developmental toxicity classes using single-species mammalian models for the rat, mouse, rabbit, as well as composite trans-species QSAR models. Because the reprotox data sets had low percentages of active chemicals and the MC4PC program is optimized for training data sets containing approximately 50% active molecules, all of the reprotox experiments assessed the impact of increasing the ratio of
Comprehensive battery of reprotox QSARs for untested chemicals
In this report, we describe the construction, optimization and validation of a comprehensive battery of QSAR models to predict seven different general reprotox classes: male and female reproductive toxicity; fetal dysmorphogenesis, functional toxicity, mortality, growth, and newborn behavioral toxicity. These reprotox QSARs are unique in that they incorporate a WOE paradigm that uses data from as many as three mammalian species (rats, mice, and rabbits), and are designed to identify
Acknowledgments
We thank Dr. Roustem Saiakhov for checking the accuracy of the chemical structures used in the experiments described in this paper and Mr. Matt Fuller for prompt technical support for MC4PC during the course of this investigation. We also thank Ms. Susan Matthews for assistance in data collection and Ms. Amie Rodgers for performing some of the validation experiments used in this study. Lastly, we thank Drs. Ruth Merkatz and Mariatta Anthony of the FDA’s Office of Women’s Health (OWH) and Dr.
References (53)
- et al.
Predicting the carcinogenic potential of pharmaceuticals in rodents using molecular structural similarity and E-state indices
Regul. Toxicol. Pharmacol.
(2003) - et al.
Evaluation of the ability of a battery of three in vitro tests to discriminate rodent carcinogens and non-carcinogens. I. Sensitivity, specificity and relative predictivity
Mutat. Res.
(2005) - et al.
Structure–activity relationship study of a diverse set of estrogen receptor ligands (I) using MultiCASE expert system
Chemosphere
(2003) - et al.
Screening of high production volume chemicals for estrogen receptor binding activity (II) by the MultiCASE expert system
Chemosphere
(2003) - et al.
A new highly specific method for predicting the carcinogenic potential of pharmaceuticals in rodents using enhanced MCASE QSAR-ES software
Regul. Toxicol. Pharmacol.
(1998) - et al.
A comprehensive model for reproductive and developmental toxicity hazard identification: I. Development of a weight of evidence QSAR database
Regul. Toxicol. Pharmacol.
(2007) - et al.
An analysis of genetic toxicity, reproductive and developmental toxicity, and carcinogenicity data: I. Identification of carcinogens using surrogate endpoints
Regul. Toxicol. Pharmacol.
(2006) - et al.
Analysis of genetic toxicity, reproductive and developmental toxicity, and carcinogenicity data: II. Identification of genotoxicants, reprotoxicants, and carcinogens using in silico methods
Regul. Toxicol. Pharmacol.
(2006) - et al.
Potential health effects of drinking water disinfection by-products using quantitative structure toxicity relationship
Toxicology
(2000) - 7th Amendment of the Cosmetics Directive: a ban on the use of live animals for testing cosmetic products and...
The utility of structure–activity relationship (SAR) models for prediction and covariate selection in developmental toxicity: comparative analysis of logistic regression and decision tree models
SAR QSAR Environ. Res.
Structure activity relationship studies of chemical mutagens and carcinogens: mechanistic investigations and prediction approaches
Chem. Rev.
The estrogen receptor relative binding affinities of 188 natural and xeno chemicals: structural diversity of ligands
Toxicol. Sci.
A computationally based identification algorithm for estrogen receptor ligands: Part 1. Predicting hER binding affinity
Toxicol. Sci.
The role of ligand flexibility in predicting biological activity: structure–activity relationships for aryl hydrocarbon, estrogen, and androgen receptor binding affinity
Environ. Toxicol. Chem.
Describing the validity of carcinogen screening tests
Br. J. Cancer
QSAR in toxicology. 4. Prediction of non-lethal mammalian toxicological endpoints, and expert systems for toxicity prediction. Quant
Struct-Act. Relat.
Use of QSARs in international decision-making frameworks to predict health effects of chemical substances
Environ. Health Perspect.
A dichotomy in the lipophilicity of natural estrogens, xenoestrogens, and phytoestrogens
Environ. Health Perspect.
Estimating the extent of the health hazard posed by high-production volume chemicals
Environ. Health Perspect.
Teratogenesis: a statistical structure–activity model
Teratogen. Carcinogen. Mutagen.
Structural determinants associated with risk of human developmental toxicity
Am. J. Obstet. Gynecol.
A QSAR model for teratogenesis
Quant. Struct.-Act. Relat.
Assessment of developmental toxicity potential of chemicals by quantitative structure-toxicity relationship models
Chemosphere
Structural determinants of developmental toxicity in hamsters
Teratology
Cited by (52)
QSAR models for predicting in vivo reproductive toxicity
2023, QSAR in Safety Evaluation and Risk AssessmentIn silico scaling and prioritization of chemical disposition and chemical toxicity of 15,145 organic chemicals
2019, Computational ToxicologyCitation Excerpt :csv) that were inserted in Excel Workbook spreadsheets. The development of the genetic toxicity, carcinogenicity, developmental toxicity, and teratogenicity QSAR models for the Multicase and Leadscope software programs has been reported in detail [58–62]. The estimates of eCT of the individual test chemicals are reported in columns LX-OR, lines 1–15145, ApT1.
Current nonclinical testing paradigms in support of safe clinical trials: An IQ Consortium DruSafe perspective
2017, Regulatory Toxicology and PharmacologyIntegrating in silico models to enhance predictivity for developmental toxicity
2016, ToxicologyCitation Excerpt :The developmental toxicity models of the rodent fetus include sub-models for effects such as dysmorphogenesis (structural and visceral birth defects), developmental toxicity (fetal growth retardation and weight decrease), and fetal survival (fetal death, post-implantation loss, and pre-implantation loss) (Table 2). The methods and data sources are described (Matthews et al., 2007a, 2007b). Developmental toxicity was modeled for study calls (e.g., dysmorphogenesis of rat), where the positive calls are trained as binary 1 and negative calls as binary 0.
Special endpoint and product specific considerations in pharmaceutical acceptable daily exposure derivation
2016, Regulatory Toxicology and PharmacologyCitation Excerpt :One should be cautious when reading-across potency as in vivo pharmacodynamics and pharmacokinetics are often compound-specific (Reichard et al., 2016, this issue). Structure-activity relationship (SAR) models can suggest potential for developmental or reproductive toxicity; however the predictive value may be limited (Arena et al., 2004; Gombar et al., 1995; Maślankiewicz et al., 2005; Matthews et al., 2007). In some cases, it can be reasonably assumed that the pharmacological target is the most sensitive endpoint, and that no adverse effects, including developmental or reproductive toxicity, would be observed below a pharmacologically effective dose.
Computational Toxicology Experience and Applications for Risk Assessment in the Pharmaceutical Industry
2013, Computational Toxicology: Methods and Applications for Risk Assessment
- ☆
This report is not an official US Food and Drug Administration guidance or policy statement. No official support or endorsement by the US Food and Drug Administration is intended or should be inferred.