Artificial intelligence and real-world data for drug and food safety - A regulatory science perspective.

In 2013, the Global Coalition for Regulatory Science Research (GCRSR) was established with members from over ten countries (www.gcrsr.net). One of the main objectives of GCRSR is to facilitate communication among global regulators on the rise of new technologies with regulatory applications through the annual conference Global Summit on Regulatory Science (GSRS). The 11th annual GSRS conference (GSRS21) focused on "Regulatory Sciences for Food/Drug Safety with Real-World Data (RWD) and Artificial Intelligence (AI)." The conference discussed current advancements in both AI and RWD approaches with a specific emphasis on how they impact regulatory sciences and how regulatory agencies across the globe are pursuing the adaptation and oversight of these technologies. There were presentations from Brazil, Canada, India, Italy, Japan, Germany, Switzerland, Singapore, the United Kingdom, and the United States. These presentations highlighted how various agencies are moving forward with these technologies by either improving the agencies' operation and/or preparing regulatory mechanisms to approve the products containing these innovations. To increase the content and discussion, the GSRS21 hosted two debate sessions on the question of "Is Regulatory Science Ready for AI?" and a workshop to showcase the analytical data tools that global regulatory agencies have been using and/or plan to apply to regulatory science. Several key topics were highlighted and discussed during the conference, such as the capabilities of AI and RWD to assist regulatory science policies for drug and food safety, the readiness of AI and data science to provide solutions for regulatory science. Discussions highlighted the need for a constant effort to evaluate emerging technologies for fit-for-purpose regulatory applications. The annual GSRS conferences offer a unique platform to facilitate discussion and collaboration across regulatory agencies, modernizing regulatory approaches, and harmonizing efforts.

Thresholds of Toxicological Concern for cosmetics-related substances: New database, thresholds, and enrichment of chemical space.

A new dataset of cosmetics-related chemicals for the Threshold of Toxicological Concern (TTC) approach has been compiled, comprising 552 chemicals with 219, 40, and 293 chemicals in Cramer Classes I, II, and III, respectively. Data were integrated and curated to create a database of No-/Lowest-Observed-Adverse-Effect Level (NOAEL/LOAEL) values, from which the final COSMOS TTC dataset was developed. Criteria for study inclusion and NOAEL decisions were defined, and rigorous quality control was performed for study details and assignment of Cramer classes. From the final COSMOS TTC dataset, human exposure thresholds of 42 and 7.9 µg/kg-bw/day were derived for Cramer Classes I and III, respectively. The size of Cramer Class II was insufficient for derivation of a TTC value. The COSMOS TTC dataset was then federated with the dataset of Munro and colleagues, previously published in 1996, after updating the latter using the quality control processes for this project. This federated dataset expands the chemical space and provides more robust thresholds. The 966 substances in the federated database comprise 245, 49 and 672 chemicals in Cramer Classes I, II and III, respectively. The corresponding TTC values of 46, 6.2 and 2.3 µg/kg-bw/day are broadly similar to those of the original Munro dataset.

New publicly available chemical query language, CSRML, to support chemotype representations for application to data mining and modeling.

Chemotypes are a new approach for representing molecules, chemical substructures and patterns, reaction rules, and reactions. Chemotypes are capable of integrating types of information beyond what is possible using current representation methods (e.g., SMARTS patterns) or reaction transformations (e.g., SMIRKS, reaction SMILES). Chemotypes are expressed in the XML-based Chemical Subgraphs and Reactions Markup Language (CSRML), and can be encoded not only with connectivity and topology but also with properties of atoms, bonds, electronic systems, or molecules. CSRML has been developed in parallel with a public set of chemotypes, i.e., the ToxPrint chemotypes, which are designed to provide excellent coverage of environmental, regulatory, and commercial-use chemical space, as well as to represent chemical patterns and properties especially relevant to various toxicity concerns. A software application, ChemoTyper has also been developed and made publicly available in order to enable chemotype searching and fingerprinting against a target structure set. The public ChemoTyper houses the ToxPrint chemotype CSRML dictionary, as well as reference implementation so that the query specifications may be adopted by other chemical structure knowledge systems. The full specifications of the XML-based CSRML standard used to express chemotypes are publicly available to facilitate and encourage the exchange of structural knowledge.

Pathways of Toxicity.

Despite wide-spread consensus on the need to transform toxicology and risk assessment in order to keep pace with technological and computational changes that have revolutionized the life sciences, there remains much work to be done to achieve the vision of toxicology based on a mechanistic foundation. To this end, a workshop was organized to explore one key aspect of this transformation - the development of Pathways of Toxicity as a key tool for hazard identification based on systems biology. Several issues were discussed in depth in the workshop: The first was the challenge of formally defining the concept of a Pathway of Toxicity (PoT), as distinct from, but complementary to, other toxicological pathway concepts such as mode of action (MoA). The workshop came up with a preliminary definition of PoT as "A molecular definition of cellular processes shown to mediate adverse outcomes of toxicants". It is further recognized that normal physiological pathways exist that maintain homeostasis and these, sufficiently perturbed, can become PoT. Second, the workshop sought to define the adequate public and commercial resources for PoT information, including data, visualization, analyses, tools, and use-cases, as well as the kinds of efforts that will be necessary to enable the creation of such a resource. Third, the workshop explored ways in which systems biology approaches could inform pathway annotation, and which resources are needed and available that can provide relevant PoT information to the diverse user communities.

Assessing the toxicity of polymeric food-contact substances.

The US Food and Drug Administration's Office of Food Additive Safety in the Center for Food Safety and Applied Nutrition conducts safety assessments of food additives, including food-contact substances such as polymeric and oligomeric materials that have the potential to migrate to food. Traditionally, little toxicity testing has been conducted on the low-molecular weight oligomeric fraction (< 1000 Da) of these food-contact substances. At lower exposures (≤ 150 µg/person/day), safety has been assessed based on the use of toxicity data on the monomeric components of these polymers as a sufficiently conservative approach for addressing the concern for genetic toxicity and carcinogenicity of the low-molecular weight oligomers (LMWOs). This paper discusses this assumption relative to the available data on these substances and their monomeric components in the context of exposures of ≤ 150 µg/person/day with emphasis on the evaluation of the potential genetic toxicity of these compounds. In most instances, data are available on either the monomers or the monomers' structural class to conservatively address the potential genetic toxicity of the LMWOs. Caveats to this generalization are also discussed. The assessment of LMWOs is important because they can be one of the primary migrants to food from a polymeric food-contact substance.

Regulatory use of computational toxicology tools and databases at the United States Food and Drug Administration's Office of Food Additive Safety.

Over 10 years ago, the Office of Food Additive Safety (OFAS) in the FDA's Center for Food Safety and Applied Nutrition implemented the formal use of structure-activity relationship analysis and quantitative structure-activity relationship (QSAR) analysis in the premarket review of food-contact substances. More recently, OFAS has implemented the use of multiple QSAR software packages and has begun investigating the use of metabolism data and metabolism predictive models in our QSAR evaluations of food-contact substances. In this article, we provide an overview of the programs used in OFAS as well as a perspective on how to apply multiple QSAR tools in the review process of a new food-contact substance.

A structural feature-based computational approach for toxicology predictions.

IMPORTANCE OF THE FIELD: Evaluation of pharmaceutical-related toxicities using quantitative structure-activity relationship (QSAR) software as decision support tools is becoming practical and is of keen interest to scientists in both product safety and discovery. QSARs can be used to predict preclinical and clinical endpoints, drug metabolism, pharmacokinetics and mechanisms responsible for toxicity. These in silico tools are of interest in supporting regulatory review processes, and priority setting in research and product development. AREAS COVERED IN THIS REVIEW: A critical assessment of the current capabilities of a new technology, the Leadscope Model Applier, is presented. Possible strengths and limitations of this technology with emphasis on the chemoinformatics method are described, and supporting literature citations date back to 1983. WHAT THE READER WILL GAIN: Insight will be gained into the Leadscope Model Applier technology for structural feature-based QSAR models and its potential capability for chemical inference if the training sets are transparently open. Currently, however, there is a lack of transparency due to the protection of the proprietary training set. TAKE HOME MESSAGE: Further research and development is needed in the creation of more stringently validated models with greater transparency and better balance between sensitivity and specificity.

Testing computational toxicology models with phytochemicals.

Computational toxicology employing quantitative structure-activity relationship (QSAR) modeling is an evidence-based predictive method being evaluated by regulatory agencies for risk assessment and scientific decision support for toxicological endpoints of interest such as rodent carcinogenicity. Computational toxicology is being tested for its usefulness to support the safety assessment of drug-related substances (e.g. active pharmaceutical ingredients, metabolites, impurities), indirect food additives, and other applied uses of value for protecting public health including safety assessment of environmental chemicals. The specific use of QSAR as a chemoinformatic tool for estimating the rodent carcinogenic potential of phytochemicals present in botanicals, herbs, and natural dietary sources is investigated here by an external validation study, which is the most stringent scientific method of measuring predictive performance. The external validation statistics for predicting rodent carcinogenicity of 43 phytochemicals, using two computational software programs evaluated at the FDA, are discussed. One software program showed very good performance for predicting non-carcinogens (high specificity), but both exhibited poor performance in predicting carcinogens (sensitivity), which is consistent with the design of the models. When predictions were considered in combination with each other rather than based on any one software, the performance for sensitivity was enhanced, However, Chi-square values indicated that the overall predictive performance decreases when using the two computational programs with this particular data set. This study suggests that complementary multiple computational toxicology software need to be carefully selected to improve global QSAR predictions for this complex toxicological endpoint.

Computational toxicology approaches at the US Food and Drug Administration.

For over a decade, the United States Food and Drug Administration (US FDA) has been engaged in the applied research, development, and evaluation of computational toxicology methods used to support the safety evaluation of a diverse set of regulated products. The basis for evaluating computational toxicology methods is multi-factorial, including the potential for increased efficiency, reduction in the numbers of animals used, lower costs, and the need to explore emerging technologies that support the goals of the US FDA's Critical Path Initiative (e.g. to make decision support information available early in the drug review process). The US FDA's efforts have been facilitated by agency-approved data-sharing agreements between government and commercial software developers. This commentary review describes former and current scientific initiatives at the agency, in the area of computational toxicology methods. In particular, toxicology-based QSAR models, ToxML databases and knowledgebases will be addressed. Notably, many of the computational toxicology tools available are commercial products - however, several are emerging as non-commercial products, which are freely-available to the public, and which will facilitate the understanding of how these programs work and avoid the "black box" paradigm. Through productive collaborations, the US FDA Center for Drug Evaluation and Research, and the Center for Food Safety and Applied Nutrition, have worked together to evaluate, develop and apply these methods to chemical toxicity endpoints of regulatory interest.

FDA toxicity databases and real-time data entry.

Structure-searchable electronic databases are valuable new tools that are assisting the FDA in its mission to promptly and efficiently review incoming submissions for regulatory approval of new food additives and food contact substances. The Center for Food Safety and Applied Nutrition's Office of Food Additive Safety (CFSAN/OFAS), in collaboration with Leadscope, Inc., is consolidating genetic toxicity data submitted in food additive petitions from the 1960s to the present day. The Center for Drug Evaluation and Research, Office of Pharmaceutical Science's Informatics and Computational Safety Analysis Staff (CDER/OPS/ICSAS) is separately gathering similar information from their submissions. Presently, these data are distributed in various locations such as paper files, microfiche, and non-standardized toxicology memoranda. The organization of the data into a consistent, searchable format will reduce paperwork, expedite the toxicology review process, and provide valuable information to industry that is currently available only to the FDA. Furthermore, by combining chemical structures with genetic toxicity information, biologically active moieties can be identified and used to develop quantitative structure-activity relationship (QSAR) modeling and testing guidelines. Additionally, chemicals devoid of toxicity data can be compared to known structures, allowing for improved safety review through the identification and analysis of structural analogs. Four database frameworks have been created: bacterial mutagenesis, in vitro chromosome aberration, in vitro mammalian mutagenesis, and in vivo micronucleus. Controlled vocabularies for these databases have been established. The four separate genetic toxicity databases are compiled into a single, structurally-searchable database for easy accessibility of the toxicity information. Beyond the genetic toxicity databases described here, additional databases for subchronic, chronic, and teratogenicity studies have been prepared.

In silico toxicological screening of natural products.

ABSTRACT This study closely examines six well-known naturally occurring dietary chemicals (estragole, pulegone, aristolochic acid I, lipoic acid, 1-octacosanol, and epicatechin) with known human exposure, chemical metabolism, and mechanism of action (MOA) using in silico screening methods. The goal of this study was to take into consideration the available information on these chemicals in terms of MOA and experimentally determined toxicological data, and compare them to the in silico predictive modeling results produced from a series of computational toxicology software. After these analyses, a consensus modeling prediction was formulated in light of the weight of evidence for each natural product. We believe this approach of examining the experimentally determined mechanistic data for a given chemical and comparing it to in silico generated predictions and data mining is a valid means to evaluating the utility of the computational software, either alone or in combination with each other. We find that consensus predictions appear to be more accurate than the use of only one or two software programs and our in silico results are in very good agreement with the experimental toxicity data for the natural products screened in this study.

Prediction of rodent carcinogenic potential of naturally occurring chemicals in the human diet using high-throughput QSAR predictive modeling.

Consistent with the U.S. Food and Drug Administration (FDA) Critical Path Initiative, predictive toxicology software programs employing quantitative structure-activity relationship (QSAR) models are currently under evaluation for regulatory risk assessment and scientific decision support for highly sensitive endpoints such as carcinogenicity, mutagenicity and reproductive toxicity. At the FDA's Center for Food Safety and Applied Nutrition's Office of Food Additive Safety and the Center for Drug Evaluation and Research's Informatics and Computational Safety Analysis Staff (ICSAS), the use of computational SAR tools for both qualitative and quantitative risk assessment applications are being developed and evaluated. One tool of current interest is MDL-QSAR predictive discriminant analysis modeling of rodent carcinogenicity, which has been previously evaluated for pharmaceutical applications by the FDA ICSAS. The study described in this paper aims to evaluate the utility of this software to estimate the carcinogenic potential of small, organic, naturally occurring chemicals found in the human diet. In addition, a group of 19 known synthetic dietary constituents that were positive in rodent carcinogenicity studies served as a control group. In the test group of naturally occurring chemicals, 101 were found to be suitable for predictive modeling using this software's discriminant analysis modeling approach. Predictions performed on these compounds were compared to published experimental evidence of each compound's carcinogenic potential. Experimental evidence included relevant toxicological studies such as rodent cancer bioassays, rodent anti-carcinogenicity studies, genotoxic studies, and the presence of chemical structural alerts. Statistical indices of predictive performance were calculated to assess the utility of the predictive modeling method. Results revealed good predictive performance using this software's rodent carcinogenicity module of over 1200 chemicals, comprised primarily of pharmaceutical, industrial and some natural products developed under an FDA-MDL cooperative research and development agreement (CRADA). The predictive performance for this group of dietary natural products and the control group was 97% sensitivity and 80% concordance. Specificity was marginal at 53%. This study finds that the in silico QSAR analysis employing this software's rodent carcinogenicity database is capable of identifying the rodent carcinogenic potential of naturally occurring organic molecules found in the human diet with a high degree of sensitivity. It is the first study to demonstrate successful QSAR predictive modeling of naturally occurring carcinogens found in the human diet using an external validation test. Further test validation of this software and expansion of the training data set for dietary chemicals will help to support the future use of such QSAR methods for screening and prioritizing the risk of dietary chemicals when actual animal data are inadequate, equivocal, or absent.