Incorporation of target safety review in drug discovery

Abstract

Ensuring the safety of new pharmaceuticals and therapies is paramount in drug discovery. Takeda’s Discovery Toxicology plays a crucial role in reducing safety-related risks by focusing on the quality of drug discovery targets. This involves identifying potential toxicities early in the development phase, enabling the mitigation of risks before they impact later stages of drug development. The Target Safety Review (TSR) initiates the Target Safety Assessment (TSA) process, providing a strategic assessment of safety concerns arising from target modulation. The TSR types range from comprehensive evaluations to simplified versions, assessing on-target and off-target effects, project background, biological information, and chemistry. Importantly, the TSR includes risk ranking and de-risking plans. Risks are ranked based on their probability and impact, enabling informed decision-making throughout the drug development process. This minireview discusses several case studies at Takeda, illustrating the importance of early risk identification. Of course, challenges remain, such as the appropriate timing of TSR creation, limited human on-target information, the need for effective risk assessment methods, incorporation of safety indicators into pharmacological studies, and addressing specific background risks in patient populations. Nonetheless, utilizing the TSR and TSA processes ensures a streamlined and safer drug development journey and provides a comprehensive approach to effectively address potential safety risks.

INTRODUCTION

Drug discovery, focused on reducing safety risks, is enhanced by modern technologies, while mechanistic toxicology improves predictions and translates findings for safer pharmaceutical applications benefiting patients. (Pognan et al., 2023) (Chen et al., 2018). At Takeda, Discovery Toxicology plays a crucial role in this process. By focusing on the quality of the drug discovery target itself, it aims to reduce safety-related risks. This involves identifying potential toxicities early in the development phase, enabling the mitigation of risks before they can impact further stages of drug development.

The Target Safety Assessment (TSA) is a comprehensive evaluative process aimed at systematically identifying, characterizing, and managing potential safety risks associated with drug targets during drug discovery and development (Gashaw et al., 2011) (Brennan, 2017). The TSA represents a multidimensional approach, incorporating in-depth biological, pharmacological, and toxicological analyses to predict and mitigate adverse effects arising from modulating specific targets. Critical components of the TSA include experimental validation of on-target effects, where the therapeutic and unintended safety risks of target modulation are actively tested in cellular or animal models. These multidirectional efforts aim to build a robust risk profile early in development, reduce late-stage failures, and ensure compliance with regulatory standards.

A foundational document that initiates the TSA is the Target Safety Review (TSR). The TSR summarizes existing information about a drug target, including its characteristics, biological functions, and disease relevance, providing the basis for the TSA. By organizing known data, the TSR helps identify potential risks and facilitates the development of mitigation strategies, such as refining drug candidates or improving project strategy. Hence, serving as the starting point for the TSA, the TSR plays a key role in guiding decisions throughout drug development.

Effective communication with project members and collective review of the TSR allows for the incorporation of safety mitigation strategies and decision criteria into project strategies. By discussing and planning exploratory tests based on evidence and hypotheses, early verification of risk is enabled. When multiple candidate targets are available, prioritizing targets with lower safety risks becomes possible, ensuring a more streamlined and safer drug development process. Given the limited disclosures from pharmaceutical companies regarding TSA efforts (Roberts, 2018) (Rudmann, 2013), this paper aims to present examples from Takeda’s TSR initiatives.

CONTENTS OF THE TSR

The TSR includes several distinct types, varying in complexity and scope to address diverse needs within drug discovery and development. Table 1 provides an overview of those types, illustrating their applications and characteristics within Takeda’s framework. While the full TSR involves a thorough and detailed assessment of potential risks, there are simplified versions tailored to specific needs, such as narrowing the scope to fewer investigation targets, prioritizing urgent deadlines, or focusing on particular safety concerns. It is important to note that this classification of TSR types is defined specifically within Takeda and may not align with general classifications used elsewhere. The key components of the full TSR in Takeda are as follows, and frequently used websites are listed in Table 2.

Table 1. Types of the TSRs.

Full TSR	Conducted at the start of the project to comprehensively evaluate safety concerns. It anticipates potential adverse events if the pharmacological action extends beyond the intended target and ranks the risks.
Mini-TSR	Focuses on specific investigations (e.g., off-target hits, past project toxicity test results) rather than being comprehensive. It may include Risk Assessment and De-Risking Plans and is designed to be prepared quickly for early decision making.
Micro-TSR	A simplified version summarizing only impactful information on one page. It is used when evaluating multiple drug discovery targets simultaneously or investigating off-target possibilities when unexpected toxicity is observed. It is employed when multiple targets are identified during phenotypic screening using annotation library compounds and also when off-target gene knockdown is observed due to oligonucleotide therapies. It does not include Risk Ranking or Risk Assessment and De-Risking Plans.
Modality Safety Review	Provided alongside the TSR for new modalities. Once created, a Modality Safety Review can be utilized for other projects, streamlining the process, and ensuring consistent safety evaluations across different initiatives.

There are various types of TSRs, ranging from comprehensive evaluations conducted at the start of a project to simplified versions focused on specific purposes. This classification of TSR types is defined specifically within Takeda.

Table 2. Frequently used websites.

Information	Website	Source
Function	Open Targets	https://www.opentargets.org/
	IUPHAR/BPS Guide to PHARMACOLOGY	https://www.guidetopharmacology.org/
	PubMed	https://pubmed.ncbi.nlm.nih.gov/
Interaction/pathway information	MetaCore	https://clarivate.com/
	IPA	https://digitalinsights.qiagen.com/
	STRING	https://string-db.org/
Protein/gene expression	UniProt	https://www.uniprot.org/
	THE HUMAN PROTEIN ATLAS	https://www.proteinatlas.org/
	Expression Atlas	https://www.ebi.ac.uk/gxa/home
	GTEx Portal	https://www.gtexportal.org/home/
	BioGPS	http://biogps.org/
Human annotation/genetics information	DISGENET	https://www.disgenet.org/search
	OMIM	https://www.omim.org/
Transgenic/knockout animal phenotypes	International Mouse Phenotyping Consortium (IMPC)	https://www.mousephenotype.org/
	Mouse Genome Informatics	http://www.informatics.jax.org/
Safety information of Drugs	Cortellis	https://clarivate.com/
	OFF-X	https://clarivate.com/
	PharmaPendium	https://pharmapendium.com/
	ChEMBL	https://www.ebi.ac.uk/chembldb/

A single website can provide multiple pieces of information, which are categorized as typical uses. For some websites, a license fee applies.

Project background

This section begins by concisely summarizing the information obtained from the project, e.g., target indication, modality, project timeline, route of administration, screening assay, targeted pharmacodynamic level, information of a prior product/compound/program, and any comorbidities identified.

Basic biological information

Instead of describing effects, this section details the profile of the target protein, e.g., functions, pathways, expression in humans/animals, homology/species differences. For human/animal expression, we use not only public databases but also our own in-house data.

On-target effects

Investigates the biological and pharmacological effects when the target protein is modulated, e.g., pathway analysis, human annotation/genetics information, transgenic/knockout animal phenotypes, safety information of control drugs/preceding products. Detailed explanations are provided for risks potentially fatal to the project, irrespective of being in vitro or in vivo. If there is conflicting information, both sides are represented fairly.

Off-target effects

The TSR primarily investigates on-target risks, however it also includes related off-target risks. This involves identifying protein groups expected to have high homology based on phylogenetic tree analysis and, if the binding sites of lead compounds are known, assessing off-target potential by checking binding pocket homology within those protein groups. Additionally, safety risks arising from the route of administration and modality are considered. For example, in cases involving new modalities such as oligonucleotide therapies or lipid nanoparticles, a Modality Safety Review is briefly prepared (Table 1) based on relevant guidance or white papers and provided as an addition to the TSR.

Chemistry

Safety risks in in vivo toxicity studies can sometimes be estimated based on physicochemical properties (Hughes et al., 2008), and Takeda's internal analysis also showed that in vivo toxicity odds increased under conditions where exposure was specified at a certain exposure level (Yukawa and Naven, 2020). If a hit compound has already been identified, risk avoidance based on structure should be considered, while also taking into account physicochemical properties that can maintain efficacy.

Risk ranking

Risks are ranked based on the probability (likelihood of “hazard or adverse events” occurring in humans, based on in vivo or clinical information and the modality used) and impact (potential for project no-go or delays if the “hazard or adverse events” materializes). Each is divided into three levels: High, Middle, and Low. These levels are detailed in Table 3. This is not a one-size-fits-all approach; the grading can vary depending on factors such as administration method and targeted disease, even with the same evidence. Proactive measures are considered when both probability and impact are ranked as High or the combination is Middle/High.

Table 3. Risk Ranking.

Probability
High	Supported by strong clinical evidence such as on market drugs or human genetic annotations (e.g., mutations, single nucleotide polymorphisms).Multiple sources of supporting evidence from human, in vivo, and in vitro studies are available.Clinical evidence or significant findings in knockout (KO) mouse models combined with multiple sources of preclinical evidence.
Middle	Suggested by in vivo experiments or strong preclinical studies, such as findings in KO mouse models or pharmacological/toxicity studies.Multiple sources of preclinical evidence, strong indications from KO mice, or clear clinical evidence.Several animal study results without definitive human data.
Low	Based on controversial or limited evidence, including in vitro mechanism-based studies, singular KO mouse findings, or isolated reports.Suggested by molecular functions with controversial or weak supporting preclinical evidence.In vitro data indicating possible in vivo implications without strong backing from other evidence types.
Impact
High	Life-threatening or severely impacts quality of life, and unmanageable.Associated with acute severe toxicity (e.g., neurotoxicity), genotoxicity leading to no-go decisions, or unmanageable toxicities.Severe and poses significant, irreversible harm.
Middle	Severe but manageable toxicity, where toxicity is monitorable and controllableImpacts quality of life similar to some existing clinical drugs’ adverse effects but is manageable with additional effort.
Low	Not life-threatening, easily monitorable, and recoverable.Minor impact on quality of life considering the indication, manageable with minimal effort.Manageable toxicities with the existence of effective treatments to cure or mitigate the effects.

Risk Ranking is based on the information provided in the TSR, grading the probability and impact of identified risks. However, it is not a one-size-fits-all approach; the grading can vary depending on factors such as administration method and target disease, even with the same evidence.

Risk assessment and de-risking plans

This section includes the addition of specific readouts in standard toxicity tests and well as a proposal of additional evaluations using fit for purpose internal and external assays and models to address and mitigate identified risks.

Safety biomarkers

This section may include organ-specific markers and the incorporation of safety indicators within in vivo nonclinical pharmacological studies to monitor and evaluate potential toxicity.

Target safety summary

As a last context of the TSR, a table named the Target Safety Summary is included. The Target Safety Summary includes the On-Target Effects with reported findings, the Risk Ranking, and the Risk Assessment and De-Risking Plans.

CASE STUDIES OF THE TSA

Based on the information provided by the TSR, various strategies can be formulated to address the identified safety concerns. These strategies are designed to mitigate risks and ensure the safe development of drug candidates (Coltman et al., 2023). The following case studies in Takeda illustrate how the TSR insights are applied to develop and implement specific safety strategies. Case studies were listed in Table 4.

Table 4. List of case studies.

	Project background	Risks	Reported findings	Probability	Impact
Case 1	Indication: Central Nervous System DisordersTarget Organ: BrainModality/MOA: Nucleic acid medicine as enzyme inhibitors	Cardiovascular risk	Reported cardiac function changes in a specific knockout mouse strain; however, another strain showed no changes, making the findings controversial	Middle	High
Case 2	Indication: Non-Alcoholic SteatohepatitisTarget Organ: LiverModality/MOA: Small molecule lipase inhibitors	Organ damage	Multiple safety concerns reported in organs other than the liver; no significant concerns in the liver	Middle	High
Case 3	Indication: Non-Alcoholic SteatohepatitisTarget Organ: LiverModality/MOA: Small molecule kinase inhibitors	Tumorigenesis	Reported liver tumors in knockout mice	Middle	High

Representative risks are presented in the form of a Target Safety Summary. Project background information is usually summarized at the beginning of the TSR and is not included in the Target Safety Summary. As for reported findings, only reports and information that are key to the TSA are listed. Strategies such as de-risking activities are described in the Results. Each case originates from a different project.

Case 1: Utilizing KO mouse insights

This case study explores central nervous system disorders targeting the brain, using nucleic acid medicine as enzyme inhibitors, with reported cardiac function changes in one knockout mouse strain while another strain showed no changes, highlighting controversial findings. Since the findings were limited to specific mouse strains and there is no information from humans, it was considered crucial to initially validate the risk identified in a KO mouse by evaluating human extrapolation. Therefore, verifying with human cells was deemed useful. The strategy focuses on evaluating cardiac function in human iPS cell-derived cardiomyocytes with reduced target gene/protein expression. Furthermore, if risks are identified in human cells, it was anticipated that discussions regarding the safety margin would inevitably ensue, making it important to conduct electrocardiogram evaluations in animal species that overlap with nucleic acid medicine as early as possible. Therefore, to more accurately estimate the risk, the strategy proposed early validation of human extrapolation using tool nucleic acids in human iPS cell-derived cardiomyocytes and early implementation of in vivo electrocardiogram evaluations.

Case 2: Modality change

This case study investigated non-alcoholic steatohepatitis with a small molecule lipase inhibitor, reporting multiple safety concerns in non-liver organs but no significant issues in the liver. The lipase had closely related family proteins with high homology. A micro-TSR investigation revealed that these homologous lipases were associated with undesirable off-target effects. Considering the inhibition mechanism and binding sites of small molecules, it was deemed challenging to ensure target lipase selectivity. Given the predicted on-target side effects in non-liver tissues, it was desired to develop a drug with as high liver selectivity as possible. Given that nucleic acid medicine would more likely to achieve high liver-selective delivery, the project strategy involved changing the modality from small molecules to nucleic acid medicine to increase lipase selectivity and liver selectivity. Even though the shift to nucleic acid medicine was expected to enhance liver selectivity, it was recommended to monitor adverse events in non-liver organs within in vivo pharmacological studies.

Case 3: Addressing proliferative changes

This case study addressed non-alcoholic steatohepatitis using a small molecule kinase inhibitor, with findings of liver tumors in knockout mice. The kinase was one of the targets identified during phenotypic screening using annotation library compounds. Based on the concentration dependency and the subsequent pharmacodynamic effects in the assay, the efficacy required persistent and strong inhibition of the target. Liver was the target organ, and tumors were reported in the livers of the target KO mice. Considering the mechanism, there was a potential for tumor development even with postnatal inhibition. Without conducting a full carcinogenicity study, it was determined that the risk of liver tumors could not be definitively excluded. Although the TSR fundamentally does not propose a no-go decision, fortunately, there were other potential targets available at that time, so de-prioritization was proposed.

CURRENT STATUS AND CHALLENGES

The timing of TSR implementation

First, it is crucial to create the TSR at the appropriate timing. Ideally, if the TSR is developed before initiating large-scale screening, the project can start with a clear understanding of on-target risks, and employ counter-assays to avoid off-target effects. On the other hand, if the TSR is created after identifying lead compounds, the probability and impact of on-target risks may instead be based on the pharmacodynamic effects and tissue distribution of the lead compounds. It is ideal to treat the TSR as a living document, updating it as the program progresses and additional data is available.

The challenge of extrapolation to humans and key considerations

In many cases, especially with innovative targets, there is limited on-target risk information available from humans. Furthermore, even if there is non-clinical information suggesting on-target risks, it is crucial to consider whether these risks can be extrapolated to humans (Namdari et al., 2021). For example, it is also necessary to consider the extrapolation of knockout mouse data to humans. This is especially important when involving mechanisms with species differences between non-clinical animals and humans, such as immune functions (Mestas and Hughes, 2004). With the recent availability of large-scale genomics data, it may be possible to use such data to consider human risks and consequences (Carss et al., 2023).

The challenge of making go/no-go decisions based on the TSR

The go/no-go decision of a project should be based on experimental validation (Potter, 2015), and not fundamentally determined by the TSR. In fact, except the Case 3, a no-go decision was never made based on the TSR. On the other hand, one of the significant challenges in the TSA is dealing with the absence of appropriate risk assessment methods and mitigation strategies. This poses critical questions: What should be done when there are no suitable methods for risk evaluation? How should long durations required for in vivo risk assessments be managed such as during carcinogenicity studies? Furthermore, if in vivo risk assessment is not feasible, should decisions be made based on in vitro data only? These complex questions emphasize the need for strategic planning and flexible approaches to handle uncertainties in the drug discovery process.

The challenge of utilizing in vivo pharmacological studies

Another challenge lies in determining the extent to which safety indicators should be incorporated into pharmacological studies. Properly incorporating safety readouts into pharmacological studies as a part of the TSA might enable early project decision-making. However, managing changes or findings caused by test substances in disease models requires precise handling to ensure accurate interpretations, as specific pharmacological animal models may be in use and the disease status and phenotype of such models is not always appropriate conditions for the safety evaluation. We think it is important to understand that the mechanisms and pathways that cause on-target toxicity are maintained in the animal model. Additionally, considering safety from the margin perspective often reveals that the exposure level in such pharmacological studies may not always be sufficient to ensure safety, complicating the interpretation of results.

Considerations regarding the human population

Finally, it may be crucial to consider specific human background risks inherent to certain patient populations, such as those assessed in a Disease Safety Review. This involves addressing comorbidities and Single Nucleotide Polymorphisms and determining whether proactive risk mitigation activities should be intensified if particular risks are identified (Dafniet and Taboureau, 2024). These further considerations related to patient backgrounds are essential to integrate safety considerations effectively into the broader research and development strategy and may need to be understood disease-specifically as a Disease Safety Review.

CONCLUSION

The TSR provides a detailed assessment of various safety-related aspects concerning the drug target. The core philosophy of the TSR is not to determine go/no-go decisions based solely on the information available, but to present project strategies from a safety perspective, enabling researchers to make informed and proactive adjustments throughout the drug discovery process. The TSR creation itself may become relatively easy with the advancement of automation. However, having toxicologists take the initiative in creating TSRs and reviewing them collectively across Discovery Toxicology can provide learning opportunities and contribute to their growth as researchers. Furthermore, it remains essential to consider risk mitigation using both internal and external assets based on the risks identified by the TSR. Moreover, it is not enough to simply compile the TSR; effective communication with project members enabling collaborative decisions regarding derisking efforts is also crucial.

Conflict of interest

Yuichiro Amano, Brandon Jeffy, Hisashi Anayama, and Jodi Goodwin are employees of Takeda Pharmaceutical Company Limited and may own company stock.

REFERENCES

•  Brennan,  R.J. (2017): Target Safety Assessment: strategies and Resources. Methods Mol. Biol., 1641, 213-228.
•  Carss,  K.J.,  Deaton,  A.M.,  Del Rio-Espinola,  A.,  Diogo,  D.,  Fielden,  M.,  Kulkarni,  D.A.,  Moggs,  J.,  Newham,  P.,  Nelson,  M.R.,  Sistare,  F.D.,  Ward,  L.D. and  Yuan,  J. (2023): Using human genetics to improve safety assessment of therapeutics. Nat. Rev. Drug Discov., 22, 145-162.
•  Chen,  J.,  Luo,  X.,  Qiu,  H.,  Mackey,  V.,  Sun,  L. and  Ouyang,  X. (2018): Drug discovery and drug marketing with the critical roles of modern administration. Am. J. Transl. Res., 10, 4302-4312.
•  Coltman,  N.J.,  Roberts,  R.A. and  Sidaway,  J.E. (2023): Data science in drug discovery safety: challenges and opportunities. Exp. Biol. Med. (Maywood), 248, 1993-2000.
•  Dafniet,  B. and  Taboureau,  O. (2024): Prediction of adverse drug reactions due to genetic predisposition using deep neural networks. Mol. Inform., 43, e202400021.
•  Gashaw,  I.,  Ellinghaus,  P.,  Sommer,  A. and  Asadullah,  K. (2011): What makes a good drug target? Drug Discov. Today, 16, 1037-1043.
•  Hughes,  J.D.,  Blagg,  J.,  Price,  D.A.,  Bailey,  S.,  Decrescenzo,  G.A.,  Devraj,  R.V.,  Ellsworth,  E.,  Fobian,  Y.M.,  Gibbs,  M.E.,  Gilles,  R.W.,  Greene,  N.,  Huang,  E.,  Krieger-Burke,  T.,  Loesel,  J.,  Wager,  T.,  Whiteley,  L. and  Zhang,  Y. (2008): Physiochemical drug properties associated with in vivo toxicological outcomes. Bioorg. Med. Chem. Lett., 18, 4872-4875.
•  Mestas,  J. and  Hughes,  C.C. (2004): Of mice and not men: differences between mouse and human immunology. J. Immunol., 172, 2731-2738.
•  Namdari,  R.,  Jones,  K.,  Chuang,  S.S.,  Van Cruchten,  S.,  Dincer,  Z.,  Downes,  N.,  Mikkelsen,  L.F.,  Harding,  J.,  Jäckel,  S.,  Jacobsen,  B.,  Kinyamu-Akunda,  J.,  Lortie,  A.,  Mhedhbi,  S.,  Mohr,  S.,  Schmitt,  M.W. and  Prior,  H. (2021): Species selection for nonclinical safety assessment of drug candidates: examples of current industry practice. Regul. Toxicol. Pharmacol., 126, 105029.
•  Pognan,  F.,  Beilmann,  M.,  Boonen,  H.C.,  Czich,  A.,  Dear,  G.,  Hewitt,  P.,  Mow,  T.,  Oinonen,  T.,  Roth,  A.,  Steger-Hartmann,  T.,  Valentin,  J.P.,  Van Goethem,  F.,  Weaver,  R.J. and  Newham,  P. (2023): The evolving role of investigative toxicology in the pharmaceutical industry. Nat. Rev. Drug Discov., 22, 317-335.
•  Potter,  W.Z. (2015): Optimizing early Go/No Go decisions in CNS drug development. Expert Rev. Clin. Pharmacol., 8, 155-157.
•  Roberts,  R.A. (2018): Understanding drug targets: no such thing as bad news. Drug Discov. Today, 23, 1925-1928.
•  Rudmann,  D.G. (2013): On-target and off-target-based toxicologic effects. Toxicol. Pathol., 41, 310-314.
•  Yukawa,  T. and  Naven,  R. (2020): Utility of Physicochemical Properties for the Prediction of Toxicological Outcomes: takeda Perspective. ACS Med. Chem. Lett., 11, 203-209.