ADMET-S Evaluation in Modern Drug Design and Discovery

Authors: Ariana Hung (1), Salman Siddiqi (2), Hadia Syed (3)

Editor: Monsurat Lawal, Ph.D.

Authors' Affiliations: (1) University of Hong Kong, (2) Willowbrook High School, (3) Urbana High School, Ijamsville, MD.

Keywords: Drug Discovery, ADMET-S, Pharmacokinetics, Drug Metabolism, Toxicity

Introduction

Drug discovery and design is a complex process that involves numerous stages: target identification, lead compound discovery, in vitro/in vivo studies, preclinical studies, clinical development, and market approval. One of the most important steps in this process is the evaluation of ADMET-S properties. ADMET-S is the abbreviation for absorption, distribution, metabolism, elimination, toxicity, and stability. These properties allow us to determine the pharmacokinetics, safety, and chemical stability of a certain drug for human use [1].

The Importance of ADMET-S in Drug Design

What is ADMET-S?

ADME began in the mid-20th century and later gained recognition due to the need to reduce the high attrition rates of drug candidates in clinical trials. The term finally emerged as a holistic approach, combining high-throughput screening technologies alongside computational models to assess ADMET-S efficiently. Today, ADMET-S (Figure 1) studies are integral to drug discovery and development worldwide, streamlining the selection of a potential drug candidate for further preclinical and clinical evaluations [2].

Consequences of Neglecting ADMET-S

Neglecting ADMET-S properties leads to drug failure, significant delays in the drug design and discovery process, and adverse effects on patients. Examples of drugs withdrawn due to poor ADMET properties include:

• Posicor (mibefradil): Withdrawn due to drug-drug interactions affecting liver metabolism [3].

• Terfenadine: An antihistamine withdrawn due to cardiac toxicity when taken with certain other drugs [4].

• Trovafloxacin: An antibiotic withdrawn due to severe liver damage [5].

• Fenfluramine and Dexfenfluramine: Appetite suppressants withdrawn due to links with pulmonary hypertension and valvular heart disease [6].

• Rofecoxib (Vioxx): An anti-inflammatory drug withdrawn due to poor toxicity, metabolism, and distribution properties, increasing the risk of heart attacks and strokes [7].

• Fialuridine (FIAU): A drug for Hepatitis B was removed due to severe mitochondrial toxicity in the liver, resulting in multiple patient deaths [8].

• Other drugs withdrawn due to hepatotoxicity include Benoxaprofen, Bromfenac, Ticrynafen, Troglitazone, Tolcapone, and Trovafloxacin.

• Temafloxacin, Nomifensine, and Remoxipride were withdrawn due to hemolytic and aplastic anemia.

Key Components of ADMET-S Evaluation

Absorption: The absorption refers to the rate and method of entry of the chemical into the bloodstream [9]. The administration of drugs occurs via ingestion, inhalation, topical, and injection. The rate of absorption depends on solubility, permeability, and structure. Drugs pass membranes by passive, facilitated, active, and endocytosis.

Distribution: After absorption, the drug is distributed via the bloodstream or by passage from cell to cell. Distribution determines the amount of drug that reaches its site of action and its effectiveness [9]. Blood flow, lipophilicity, binding to tissues, and the size of the molecule can all influence distribution.

Metabolism: The process by which drugs undergo biotransformation in organs for excretion is known as metabolism. Drugs are made more water-soluble to be excreted through the feces or urine [9].

Excretion: The rate or the speed at which a drug is eliminated into the feces, urine, lungs, or through the process of sweating [9]. There are various factors influencing which route of excretion might occur, and the primary factor is the size of the molecule.

Toxicity: It is the measure of the harm a particular drug can cause, which could be determined through its therapeutic index or median lethal dose [9].

Stability: The resistance of a drug molecule to any change in its integrity or therapeutic potency with time is termed chemical stability [10].

Figure 1: ADMET-S at the computational drug design level.

Challenges in ADMET-S Evaluation and the Role of Computational Approaches

One of the primary challenges lies in the complexity and variability of biological systems. ADMET properties are influenced by numerous factors, including genetic diversity, disease states, and drug interactions, making it difficult to predict how a compound will behave in humans based solely on laboratory or animal studies. Additionally, traditional experimental methods for ADMET testing are often time-consuming, expensive, and resource-intensive. These limitations can lead to late-stage failures during clinical trials, where unforeseen ADMET issues emerge, resulting in significant financial losses and delays in bringing new drugs to market [11].

Another major challenge is the balance between sensitivity and specificity in ADMET evaluations. For instance, toxicity assessments must be highly sensitive to avoid exposing patients to harmful compounds, but overly stringent criteria may unnecessarily eliminate potentially viable drug candidates [12]. Similarly, accurately predicting metabolic pathways and excretion rates requires detailed knowledge of enzyme-substrate interactions, which can vary significantly across populations. Furthermore, non-standardized protocols and differences in experimental conditions across laboratories can lead to inconsistent results, complicating the interpretation of ADMET data. These challenges highlight the need for more efficient, accurate, and scalable approaches to ADMET-S evaluation.

Computational approaches have emerged as powerful tools to address these challenges by enabling faster, cost-effective, and more predictive ADMET assessments [11]. Techniques such as molecular docking, quantitative structure-activity relationship (QSAR) modeling, and machine learning algorithms allow researchers to simulate and predict ADMET properties at an early stage of drug discovery. For example, QSAR models use statistical correlations between chemical structures and biological activities to predict absorption or toxicity profiles, reducing reliance on experimental data. Similarly, machine learning models trained on large datasets of known compounds can identify patterns and relationships that might not be apparent through traditional methods. Computational tools also facilitate high-throughput virtual screening, enabling the rapid evaluation of thousands of compounds for desirable ADMET properties before committing to costly experimental validation. By integrating computational approaches into the drug discovery pipeline, researchers can prioritize safer and more viable drug candidates, reduce late-stage attrition rates, and accelerate the development of effective therapeutics [11].

Computational Tools and Software for ADMET and Drug Stability Prediction

The accurate prediction of ADMET properties and drug stability is critical for accelerating drug discovery and minimizing late-stage failures. To address these challenges, a variety of computational tools and software platforms have been developed, leveraging advances in cheminformatics, machine learning, and molecular modeling. These tools enable researchers to predict ADMET profiles and assess drug stability at an early stage, thereby optimizing the selection of viable drug candidates.

One widely used platform is ADMETlab [13], a comprehensive web-based tool designed for ADMET property prediction. ADMETLab integrates a vast collection of curated datasets and advanced machine learning models to predict over 30 ADMET-related endpoints, including solubility, permeability, metabolic stability, and toxicity. It also provides user-friendly visualization tools to interpret results, making it accessible to both computational and non-computational scientists. The platform’s ability to handle large-scale virtual screening and its high predictive accuracy make it an invaluable resource for drug discovery teams aiming to prioritize compounds with favorable ADMET profiles.

In addition to ADMETLab, several other computational tools are commonly employed in this domain. For instance, SwissADME [14] is a free, web-based tool that predicts key physicochemical and pharmacokinetic properties, such as lipophilicity, water solubility, and blood-brain barrier penetration. Its simplicity and rapid processing capabilities make it ideal for quick assessments during the early stages of drug design. Similarly, pkCSM [15] uses graph-based signatures to predict pharmacokinetic properties and toxicity endpoints, offering insights into how structural features influence ADMET behavior. Other software, including ToxinPred, ProTox, Mcule, and Toxtree, focuses specifically on predicting toxicological outcomes, aiding in the identification of potential safety concerns before experimental validation.

For drug stability prediction, tools like StarDrop™ and Schrodinger Suite [16] provide specialized modules to evaluate chemical stability, degradation pathways, and metabolic liabilities. StarDrop’s probabilistic scoring system helps identify compounds with optimal stability and ADMET characteristics, while Schrodinger’s suite combines molecular dynamics simulations and quantum mechanics [17] to analyze the stability of drug molecules under various conditions.

These computational tools collectively enhance the efficiency and reliability of ADMET and drug stability evaluations. By integrating experimental data with predictive algorithms, they reduce the need for costly and time-consuming laboratory tests, enabling researchers to make informed decisions earlier in the drug discovery process. As the field continues to evolve, the integration of AI-driven approaches and larger, more diverse datasets will further improve the accuracy and applicability of these tools, ultimately contributing to safer and more effective therapeutics.

 

Impact of ADMET-S Failure on the Pharmaceutical Industry and Public Health

The failure of ADMET-related assessments at any stage of the drug development pipeline can have far-reaching consequences for both the pharmaceutical industry and public health.

Drug development is an expensive and time-consuming process, with estimates suggesting that it takes over a decade and billions of dollars to bring a new drug to market. A significant proportion of drug candidates fail during clinical trials due to poor ADMET properties, such as inadequate absorption, rapid metabolism, or unforeseen toxicity. These failures not only result in substantial financial losses but also delay the availability of potentially life-saving therapies. The inability to predict ADMET issues early in the drug discovery pipeline highlights inefficiencies in current screening methods, leading to wasted resources and missed opportunities for innovation within the pharmaceutical sector.

From a public health perspective, ADMET-S failures can exacerbate unmet medical needs by slowing down the introduction of new treatments. For patients suffering from chronic or life-threatening conditions, delays in drug approval can mean prolonged suffering or even premature death. Moreover, when drugs with poor ADMET profiles reach the market despite suboptimal testing, they may pose serious risks to patient safety, resulting in adverse drug reactions (ADRs) or recalls. Such incidents erode public trust in the healthcare system and regulatory agencies, making it harder to promote adherence to prescribed therapies. Ensuring robust ADMET evaluations is therefore critical not only for safeguarding patient health but also for maintaining confidence in pharmaceutical advancements.

To mitigate these challenges, the industry must adopt innovative strategies to enhance ADMET-S prediction accuracy, such as integrating artificial intelligence (AI), machine learning models, and advanced computational tools into the drug discovery process. By identifying potential ADMET issues earlier, researchers can prioritize safer and more viable drug candidates, reducing late-stage attrition rates and improving overall success rates. Furthermore, fostering collaboration between academia, industry, and regulatory bodies can lead to standardized methodologies and shared datasets, accelerating progress toward safer and more effective medicines. Addressing ADMET-S failures proactively will not only benefit the pharmaceutical industry by optimizing resource allocation but also ensure timely access to high-quality therapeutics, ultimately advancing global public health outcomes.

Future Trends in ADMET-S Evaluation

The evaluation of ADMET-S properties is poised to undergo transformative advancements driven by the integration of cutting-edge technologies and innovative methodologies. These trends are expected to address current limitations, improve prediction accuracy, and accelerate drug discovery while ensuring safer and more stable therapeutics.

One of the most significant future trends is the increasing adoption of artificial intelligence (AI) and machine learning (ML) in ADMET-S evaluation [11]. AI-driven models are becoming more sophisticated, enabling the analysis of vast datasets to identify complex patterns and relationships between molecular structures and ADMET properties. For instance, deep learning algorithms can predict drug stability (S) by simulating degradation pathways and identifying potential metabolic liabilities with unprecedented precision. Additionally, ML models trained on diverse experimental data can account for population-specific variations in ADMET profiles, such as genetic polymorphisms affecting drug metabolism. These advancements will allow researchers to prioritize compounds with optimal stability and safety earlier in the drug discovery pipeline, reducing late-stage attrition rates.

Another emerging trend is the development of multi-scale modeling approaches that integrate molecular, cellular, and organism-level data to provide a holistic view of ADMET-S properties. Techniques such as systems pharmacology and physiologically-based pharmacokinetic (PBPK) modeling simulate how drugs interact with biological systems across different scales, from molecular binding events to whole-body distribution. These models are particularly valuable for predicting drug stability under various physiological conditions and understanding the impact of co-administered drugs or disease states on ADMET profiles. Furthermore, the incorporation of quantum mechanics/molecular mechanics (QM/MM) simulations will enhance the accuracy of predictions related to metabolic stability and chemical degradation pathways.

A third trend is the growing emphasis on personalized medicine in ADMET-S evaluation. Advances in genomics, proteomics, and metabolomics are enabling the identification of biomarkers that influence individual responses to drugs. Computational tools that integrate multi-omics data with ADMET predictions will facilitate the design of therapies tailored to specific patient populations. For example, predicting drug stability and toxicity in individuals with unique metabolic profiles will help minimize adverse drug reactions (ADRs) and optimize therapeutic outcomes. This shift toward personalized ADMET-S assessments aligns with the broader goal of precision medicine, ensuring that drugs are both effective and safe for diverse patient groups.

Finally, the rise of cloud computing and collaborative platforms is set to revolutionize ADMET-S evaluation by enabling real-time data sharing and analysis across global research teams. Platforms like ADMETlab and other cloud-based tools will allow researchers to access large-scale datasets, standardized protocols, and advanced computational resources without the need for extensive local infrastructure. This democratization of ADMET-S tools will foster collaboration between academia, industry, and regulatory agencies, accelerating innovation and improving the reliability of predictions. Moreover, the integration of blockchain technology may enhance data integrity and transparency, ensuring the reproducibility of ADMET-S evaluations.

In conclusion, the future of ADMET-S evaluation lies in the convergence of AI, multi-scale modeling, personalized medicine, and collaborative technologies. These trends will not only improve the accuracy and efficiency of ADMET-S predictions but also pave the way for safer, more stable, and patient-centric therapeutics. By addressing current challenges and leveraging these advancements, the pharmaceutical industry can significantly reduce drug development costs and timelines while enhancing public health outcomes.

Conclusion

The study of ADMET-S properties plays a crucial role in determining the viability of new drugs and serves as the basis for new drug discovery. Advances in machine learning and personalized medicine drive the need for precise ADMET-S evaluations tailored to individual patient profiles, potentially revolutionizing treatment outcomes. These innovations paint a bright picture of the future of medicine.

Glossary:

1. ADMET-S: Absorption, Distribution, Metabolism, Elimination, Toxicity, and Stability; refers to critical pharmacokinetic and pharmacodynamic properties evaluated in drug development.

2. Pharmacokinetics: The study of the absorption, distribution, metabolism, and excretion of drugs in the body over time.

3. Therapeutic Index: The ratio between the toxic dose and the therapeutic dose of a drug, used to measure its safety margin.

4. High-Throughput Screening (HTS): A method for scientific experimentation that uses automation to quickly conduct millions of chemical, genetic, or pharmacological tests.

5. Quantitative Structure-Activity Relationship (QSAR): A computational approach that predicts the activity of chemical compounds based on their molecular structure.

6. Physiologically Based Pharmacokinetic (PBPK) Modeling: A mathematical modeling technique that predicts the absorption, distribution, metabolism, and excretion of chemicals in humans and animals.

7. In vitro: Experiments conducted outside of living organisms, typically in a controlled lab environment, such as in test tubes or petri dishes.

8. In vivo: Experiments conducted within living organisms, such as in animals or humans.

9. Omics Technologies: Techniques such as genomics, proteomics, and metabolomics that study biological systems comprehensively at the molecular level.

10. Bioinformatics: The application of computational tools and techniques to analyze biological data.

11. Machine Learning (ML): A subset of artificial intelligence where computer algorithms learn patterns from data to make predictions or decisions.

12. Drug Metabolism: The chemical alteration of a drug by the body, primarily in the liver.

13. Therapeutic Potency: The ability of a drug to produce a desired effect at a specific dose.

14. Humanized Models: Experimental models, such as animals, modified to carry human genes, cells, or tissues to simulate human biological processes more accurately.

References

1. Huggins, D.J., Baell, J., Brennan, P.E., Burgin, A., Scott, D.E.: The benefits of translating biomedical research at drug discovery institutes. Nat. Rev. Drug Discov. (2024). https://doi.org/10.1038/d41573-024-00142-z

2. Baillie, T.A.: Metabolism and Toxicity of Drugs. Two Decades of Progress in Industrial Drug Metabolism. Chem. Res. Toxicol. 21, 129–137 (2008). https://doi.org/10.1021/tx7002273

3. Bradbury, J.: Posicor withdrawn voluntarily from market by Roche. The Lancet. 351, 1791 (1998). https://doi.org/10.1016/S0140-6736(05)78757-2

4. Woosley, R.L., Chen, Y., Freiman, J.P., Gillis, R.A.: Mechanism of the cardiotoxic actions of terfenadine. JAMA. 269, 1532–1536 (1993)

5. Giustarini, G., Vrisekoop, N., Kruijssen, L., Wagenaar, L., van Staveren, S., van Roest, M., Bleumink, R., Bol-Schoenmakers, M., Weaver, R.J., Koenderman, L., Smit, J., Pieters, R.: Trovafloxacin-Induced Liver Injury: Lack in Regulation of Inflammation by Inhibition of Nucleotide Release and Neutrophil Movement. Toxicol. Sci. 167, 385–396 (2019). https://doi.org/10.1093/toxsci/kfy244

6. Vivero, L.E., Anderson, P.O., Clark, R.F.: A close look at fenfluramine and dexfenfluramine. J. Emerg. Med. 16, 197–205 (1998). https://doi.org/10.1016/S0736-4679(97)00289-8

7. Sibbald, B.: Rofecoxib (Vioxx) voluntarily withdrawn from market. CMAJ Can. Med. Assoc. J. J. Assoc. Medicale Can. 171, 1027–1028 (2004). https://doi.org/10.1503/cmaj.1041606

8. McKenzie, R., Fried, M.W., Sallie, R., Conjeevaram, H., Di Bisceglie, A.M., Park, Y., Savarese, B., Kleiner, D., Tsokos, M., Luciano, C.: Hepatic failure and lactic acidosis due to fialuridine (FIAU), an investigational nucleoside analogue for chronic hepatitis B. N. Engl. J. Med. 333, 1099–1105 (1995). https://doi.org/10.1056/NEJM199510263331702

9. Ferreira, L.L., Andricopulo, A.D.: ADMET modeling approaches in drug discovery. Drug Discov. Today. 24, 1157–1165 (2019)

10. Wong, A.W., Datla, A.: 13 - Assay and Stability Testing. In: Ahuja, S. and Dong, M.W. (eds.) Separation Science and Technology. pp. 335–358. Academic Press (2005)

11. Wu, F., Zhou, Y., Li, L., Shen, X., Chen, G., Wang, X., Liang, X., Tan, M., Huang, Z.: Computational Approaches in Preclinical Studies on Drug Discovery and Development. Front. Chem. 8, 726 (2020). https://doi.org/10.3389/fchem.2020.00726

12. Sun, D., Gao, W., Hu, H., Zhou, S.: Why 90% of clinical drug development fails and how to improve it? Acta Pharm. Sin. B. 12, 3049–3062 (2022). https://doi.org/10.1016/j.apsb.2022.02.002

13. Fu, L., Shi, S., Yi, J., Wang, N., He, Y., Wu, Z., Peng, J., Deng, Y., Wang, W., Wu, C.: ADMETlab 3.0: an updated comprehensive online ADMET prediction platform enhanced with broader coverage, improved performance, API functionality and decision support. Nucleic Acids Res. 52, W422–W431 (2024)

14. Daina, A., Michielin, O., Zoete, V.: SwissADME: a free web tool to evaluate pharmacokinetics, drug-likeness and medicinal chemistry friendliness of small molecules. Sci. Rep. 7, 1–13 (2017)

15. Pires, D.E.V., Blundell, T.L., Ascher, D.B.: pkCSM: Predicting Small-Molecule Pharmacokinetic and Toxicity Properties Using Graph-Based Signatures. J. Med. Chem. 58, 4066–4072 (2015). https://doi.org/10.1021/acs.jmedchem.5b00104

16. Schrödinger Suite, https://www.schrodinger.com/, https://www.schrodinger.com/citations

17. Rodrigues, R.P., Andrade, S.F., Mantoani, S.P., Eifler-Lima, V.L., Silva, V.B., Kawano, D.F.: Using free computational resources to illustrate the drug design process in an undergraduate medicinal chemistry course. J. Chem. Educ. 92, 827–835 (2015)