The control of blood pressure is complex, but vascular ion channels play a key role (Tykocki et al., 2017). Estimates suggest 20–30% of commonly prescribed medications can influence blood pressure; here we investigate the role of vascular ion channels in drug-induced blood pressure changes, offering the potential to establish a screen early in drug discovery.Sixteen marketed drugs known to affect blood pressure were tested against a panel of vascular ion channels: Kv1.1, Kv2.1, Kv7.2/7.3, Kv7.3/7.5, CaV1.2, CaV3.2, Kir2.1. Drugs were tested in vitro by automated patch clamp (QPatch II, Sophion Biosciences). Drugs known to reduce blood pressure, like nifedipine and verapamil, showed activity at several ion channels, most notable vascular calcium channels. Clozapine which also changes blood pressure, showed activity at multiple vascular ion channels including Kv2.1 and Kv7.3/7.5 with IC50s of 13.0 and 42.1 suggesting a possible mechanism of action for clozapine. The data suggest that activity at ion channels is a potential mechanism for the effect of drugs on blood pressure. Thus, when testing unknown compounds as part of a screening strategy, activities at any of these ion channels may implicate possible effect on blood pressure. However, a negative result at these vascular ion channels cannot rule out the potential for effects on blood pressure by non-ion channel mediated mechanisms. Future work will build the panel by examining other relevant ion channels such as the KATP and K2P channels.

Targeted protein degraders (TPDs) that recruit the E3 ligase cereblon (CRBN) include two innovative drug classes: molecular glue degraders and proteolysis targeting chimeras. These TPDs have shown great promise in addressing difficult drug targets in oncologic and non-oncologic diseases; however, as well as inducing proteosomal degradation of their therapeutic target, they can also lead to degradation of unintended CRBN neosubstrates. This is a major safety consideration because many CRBN-recruiting TPDs are structurally related to thalidomide and other immunomodulatory imide drugs that are known to be teratogenic. The teratogenic effect of immunomodulatory imide drugs is due in part to their induction of CRBN-mediated neosubstrate degradation. Therefore, there is a need for a scientific consensus on rigorous, consistent and effective methods to assess the safety of CRBN-recruiting TPDs. Here, we provide an overview of the endogenous functions and substrates of CRBN as well as of the role of CRBN as the effector of immunomodulatory imide drugs, next-generation molecular glue degraders and proteolysis targeting chimeras. We discuss how degradation of unintended 'off-target' CRBN neosubstrates could potentially cause toxicity or safety liabilities in multiple organ systems as well as induce teratogenic effects. We outline key safety considerations for the development of CRBN-recruiting TPDs and suggest best practices for monitoring on-target versus off-target degradation.

Seizure liability remains a key risk for drug development and following chemical exposure. Advances in induced pluripotent stem cell (iPSC) biology and in vitro detection methodologies such as multielectrode array (MEA) offer an opportunity for a new paradigm in compound screening. Thus, a coordinated effort is needed to determine the methods, models, and parameters that are optimal for larger scale nonclinical assessment for seizure liability. Here we present the results of a multi-laboratory MEA assessment of seizurogenic detection in both human iPSC-derived neural cultures and primary rat cortical neurons. We evaluated the impact of 10 pro-convulsant compounds and 3 negative controls on spontaneous electrical activity in both types of neurons utilizing MEA. The work was conducted at 7 different laboratories across 3 continents. Both rat and human models showed similar changes to common metrics such as mean of interspike distance (mean ISI) and median burst rate across facilities and across all compounds, with more consistency in the human model. Regarding differences between compounds, seizurogenic compounds caused the largest changes in MEA profile and the parameter with the most commonality was mean ISI, which was decreased in all tested compounds except picrotoxin. Overall, all laboratories generated MEA data indicative of a seizurogenic phenotype but the human model was more consistent across sites, despite some variation in protocols. Lessons learned from this work include having a clear aim before choosing a model, understanding and characterizing the model in the actual test site, and carefully considering the most appropriate timepoints to assess seizure liability.

New approach methodologies (NAMs) are increasingly recognised as central to modernising safety assessment in medicines development, offering the potential for improved human relevance, mechanistic insight, and reduced reliance on animal testing. Despite substantial scientific progress and growing regulatory encouragement in their use, the routine integration of NAMs into regulatory decision-making remains inconsistent, even though they are widely used internally across discovery and development. This article brings together several lines of evidence and opinions to explore why uptake continues to lag behind capability, including insights from a cross-sector workshop convened by the Medicines and Healthcare products Regulatory Agency (MHRA), the Association of the British Pharmaceutical Industry (ABPI) and the UK National Centre for the Replacement, Refinement and Reduction of Animals in Research (NC3Rs). For many applications, the key constraints are no longer scientific, but relate to confidence, clarity of expectations, and implementation, although important scientific challenges remain for complex endpoints such as chronic toxicity. Regulatory frameworks are increasingly supportive, with growing evidence of confidence and alignment on weight-of-evidence approaches incorporating NAMs, yet residual uncertainty and limited practical experience in their application means that questions regarding context-of-use, validation, and regulatory acceptability persist. The next phase requires a shift from technology development to practical implementation, including clearer guidance and increased transparency through the sharing of data and case-studies. Building consensus and confidence across the scientific community will be critical to normalising the use of NAMs and realising their potential in medicines development.

Central nervous system (CNS) liabilities (e.g. seizures) remain a significant cause of drug development attrition, yet they are typically not identified until late-stage in vivo toxicology studies where adverse findings can result in costly delays or project termination. Further, ~25% of clinical trial failures are due to safety concerns that were undetected in earlier in vivo studies. This emphasises the urgent need for predictive, human-relevant in vitro assays that can be implemented earlier in the drug development process to assess CNS safety risks. The cardiovascular safety field has benefitted from early-stage screening using human ion channel assays and human induced pluripotent stem cell (hiPSC)-cardiomyocytes. Similarly, we developed two complementary in vitro CNS safety assays: 1) an ion channel panel evaluated via automated electrophysiology, and 2) an hiPSC-neuronal model analyzed using microelectrode array (MEA) to measure electrical activity in response to known seizurogenic compounds. In addition to seizure risk, sedation is a common adverse finding in preclinical studies. To address this, we aimed to validate our in vitro CNS assay for predicting sedative compound effects. GABAergic signalling plays a crucial role in the sleep-wake cycle, with GABAA receptor activation resulting in sedation. We tested 14 sedative compounds (targeting both GABA and non-GABA pathways) and 5 non-sedative controls in the hiPSC-neuronal model. MEA analysis showed that GABAA agonists (muscimol, etomidate, ethanol) and certain ion channel antagonists (valbenazine, dextromethorphan, primidone) reduced neural firing and burst activity, indicating sedative effects. Negative control compounds produced no significant changes. Of the 19 known compounds screened, 85% were accurately classified. Automated patch-clamp electrophysiology further confirmed these findings by testing compound interactions with GABAA and other CNS ion channels. These results support the use of our in vitro CNS models for detecting sedative and seizurogenic effects, thus enhancing translational relevance. These platforms offer valuable human-based tools for early hazard identification and prioritisation of drug candidates, as well as interpreting CNS-related findings from in vivo studies thus complying with the principles of 3Rs to minimise and replace the use of animals in scientific testing. Ongoing work is focused on expanding this model to evaluate additional sedation-related ion channels.

Seizure liability remains a significant cause of attrition throughout drug development both in pre-clinical and clinical studies. This emphasizes the need for improved methodologies to detect seizure liability prior to in vivo toxicology studies, ideally with reduced reliance on animals and better translation to humans. Much like the Comprehensive in vitro Proarrhythmia Assay (CiPA) which is now widely accepted for early assessment of cardiovascular safety, we have developed an approach utilizing hiPSC-neuronal cell microelectrode array (MEA) and ion channel screening for early seizure prediction. In our MEA assay, seizurogenic compounds were identified correctly with high predictivity, and correlations were observed between the in vitro and clinical exposures of many therapies known to cause seizure. We have used these assays in the early phase of nonclinical testing, and successfully de-risked and prioritized a chemical series. For example, after testing a number of compounds, one was identified with low seizure risk compared to the others in the series – this compound had distinct structural features. In another study of compounds undergoing nonclinical testing, exposures that caused no CNS signs or convulsions in rats, aligned with the results of the MEA study. Conversely, where convulsions were reported in rats, seizurogenic responses were present in the MEA study at comparable concentrations. Since these studies use human derived cells, they can be used to determine the human relevance of seizures observed in nonclinical studies. For example, nonclinical testing of a compound caused convulsions only in dogs. Testing a range of metabolites in the MEA assay revealed only the dog-specific metabolite caused seizurogenic phenotype. In addition, screening this metabolite against a panel of ion channel targets revealed a hit, providing mechanistic insight and also the opportunity to redesign the compound to eliminate the liability. Collectively, these studies demonstrate the utility of this approach for early seizure prediction to provide mechanistic information, early de-risking, and support optimal drug design using human in vitro models.

The growing impact of large language models (LLMs), such as ChatGPT, prompts questions about the reliability of their application in public health. We compared drug toxicity assessments by GPT-4 for liver, heart, and kidney against expert assessments using US Food and Drug Administration (FDA) drug-labeling documents. Two approaches were assessed: a ‘General prompt’, mimicking the conversational style used by the general public, and an ‘Expert prompt’ engineered to represent an approach of an expert. The Expert prompt achieved higher accuracy (64–75%) compared with the General prompt (48–72%), but the overall performance was moderate, indicating that caution is needed when using GPT-4 for public health. To improve reliability, an advanced framework ,such as Retrieval Augmented Generation (RAG), might be required to leverage knowledge embedded in GPT-4.

There are many different approaches to drug discovery in academia, some of which are based broadly on the industrial model of discovering novel targets and then conducting screening within academic drug discovery centres to identify hit molecules. Here we describe our approach to drug discovery, which makes more efficient use of the capabilities and resources of the different stakeholders. Specifically, we have created a large portfolio of drug projects and conducted small amounts of derisking work to ensure projects are investment ready. In this feature we will describe this model, including its limitations and advantages, since we believe the ideas and concepts will be of interest to other academic institutions and consortia.

Drug-induced kidney injury (DIKI) is a frequently reported adverse event, associated with acute kidney injury, chronic kidney disease, and end-stage renal failure. Prospective cohort studies on acute injuries suggest a frequency of around 14%-26% in adult populations and a significant concern in pediatrics with a frequency of 16% being attributed to a drug. In drug discovery and development, renal injury accounts for 8 and 9% of preclinical and clinical failures, respectively, impacting multiple therapeutic areas. Currently, the standard biomarkers for identifying DIKI are serum creatinine and blood urea nitrogen. However, both markers lack the sensitivity and specificity to detect nephrotoxicity prior to a significant loss of renal function. Consequently, there is a pressing need for the development of alternative methods to reliably predict drug-induced kidney injury (DIKI) in early drug discovery. In this article, we discuss various aspects of DIKI and how it is assessed in preclinical models and in the clinical setting, including the challenges posed by translating animal data to humans. We then examine the urinary biomarkers accepted by both the US Food and Drug Administration (FDA) and the European Medicines Agency for monitoring DIKI in preclinical studies and on a case-by-case basis in clinical trials. We also review new approach methodologies (NAMs) and how they may assist in developing novel biomarkers for DIKI that can be used earlier in drug discovery and development.

Much progress has been made in reducing and refining animal use in toxicology testing, but progress in the use of new approach methodologies (NAMs) to replace animals is disapbut societal, regulatory and political barriers to their implementation remain. Change requires vision, starting with imagining a future where we are successful. Specifically, this would comprise the registration of safe and effective medicines without animal tests. How do we achieve this vision? Thinking differently, in silico methods could be used to provide a detailed assessment of target- and modality-related toxicological risks, coupled with modelling of exposure. In vitro NAMs such as microphysiological systems, microelectrode array and ion channel panels could then be employed to address hypothetical risks. Finally, the safety of first time in human trials could be assessed and assured using circulating nanobots that measure conventional clinical pathology parameters alongside new biomarkers such as circulating tissue DNA. This may seem the stuff of fantasy, but imagination is key to shaping a better future and all change starts with a vision, however far-fetched it may seem today.

Drug -induced renal injury (DIRI) causes >1.5 million adverse events annually in the USA alone. Although standard biomarkers exist for DIRI, they lack the sensitivity or speci ficity to detect nephrotoxicity before the signi ficant loss of renal function. In this study, we describe the creation of DIRIL - a list of drugs associated with DIRI and nephrotoxicity - from two literature datasets with DIRI annotation, con firmed using FDA drug labeling. DIRIL comprises 317 orally administered drugs covering all 14 anatomical, therapeutic and chemical (ATC) classi fication categories. Of the 317 drugs, 171 were DIRI-positive and 146 were DIRI-negative. DIRIL will be a relevant and invaluable resource for discovery of new approach methods (NAMs) to predict the occurrence and possible severity of DIRI earlier in drug development.

The Ames assay is required by the regulatory agencies worldwide to assess the mutagenic potential risk of consumer products. As well as this in vitro assay, in silico approaches have been widely used to predict Ames test results as outlined in the International Council for Harmonization (ICH) guidelines. Building on this in silico approach, here we describe DeepAmes, a high performance and robust model developed with a novel deep learning (DL) approach for potential utility in regulatory science. DeepAmes was developed with a large and consistent Ames dataset (>10,000 compounds) and was compared with other five standard Machine Learning (ML) methods. Using a test set of 1,543 compounds, DeepAmes was the best performer in predicting the outcome of Ames assay. In addition, DeepAmes yielded the best and most stable performance up to when compounds were >30% outside of the applicability domain (AD). Regarding the potential for regulatory application, a revised version of DeepAmes with a much-improved sensitivity of 0.87 from 0.47. In conclusion, DeepAmes provides a DL-powered Ames test predictive model for predicting the results of Ames tests; with its defined AD and clear context of use, DeepAmes has potential for utility in regulatory application.

Motivation: Antibiotic resistance presents a formidable global challenge to public health and the environment. While considerable endeavors have been dedicated to identify antibiotic resistance genes (ARGs) for assessing the threat of antibiotic resistance, recent extensive investigations using metagenomic and metatranscriptomic approaches have unveiled a noteworthy concern. A significant fraction of proteins defies annotation through conventional sequence similarity-based methods, an issue that extends to ARGs, potentially leading to their under-recognition due to dissimilarities at the sequence level. Results: Herein, we proposed an Artificial Intelligence-powered ARG identification framework using a pretrained large protein language model, enabling ARG identification and resistance category classification simultaneously. The proposed PLM-ARG was developed based on the most comprehensive ARG and related resistance category information (>28K ARGs and associated 29 resistance categories), yielding Matthew's correlation coefficients (MCCs) of 0.98360.001 by using a 5-fold cross-validation strategy. Furthermore, the PLM-ARG model was verified using an independent validation set and achieved an MCC of 0.838, outperforming other publicly available ARG prediction tools with an improvement range of 51.8%-107.9%. Moreover, the utility of the proposed PLM-ARG model was demonstrated by annotating resistance in the UniProt database and evaluating the impact of ARGs on the Earth's environmental microbiota.

Animal studies are unavoidable in evaluating chemical and drug safety. Generative Adversarial Networks (GANs) can generate synthetic animal data by learning from the legacy animal study results, thus may serve as an alternative approach to assess untested chemicals. AnimalGAN, a GAN method to simulate 38 rat clinical pathology measures, was developed with significant robustness even for the drugs that vary significantly from these used during training, both in terms of chemical structure, drug class, and the year of FDA approval. AnimalGAN showed comparable results in hepatotoxicity assessment as using the real animal data and outperformed 12 conventional quantitative structure-activity relationship approaches. Using AnimalGAN, a virtual experiment of 100,000 rats ranked hepatotoxicity of three structurally similar drugs in a similar trend that has been observed in human population. AnimalGAN represented a significant step with artificial intelligence towards the global effort in replacement, reduction, and refinement (3Rs) of animal use.