Modernizing Preclinical Drug Development: The Role of New Approach Methodologies (NAMs)

In April 2025, the U.S. Food and Drug Administration (FDA) announced that it would be involved in the development process of monoclonal antibodies and other drugs.Gradually replace animal experiments with other tools, which will be immediately implemented in the new drug clinical trial application (IND),And encourage the inclusion of NAMs data.

Advanced In Vitro Experiments:Such as 3D cell culture, Organ-on-Chip, Microphysiological Systems (MPS), and Organoids, which can better simulate the human physiological and pathological environment.

Mechanism Modeling and Artificial Intelligence:Such as Physiologically-Based Pharmacokinetic (PBPK), Quantitative Systems Pharmacology (QSP), Quantitative Systems Toxicology (QST) models, immunogenicity prediction, Digital Twin, etc., predicting human pharmacokinetics, efficacy, dosage, etc., based on in vitro/computational data.

Increase the Success Rate of New Drug Development

Significantly reduce the need for animal testing, lower R&D costs, and shorten drug time-to-market.

PK Prediction for Small Molecule and Monoclonal Antibody Drugs:The research team constructed an "a priori" PBPK model using data from in vitro experiments and physicochemical properties of compounds. Based solely on in vitro data and existing knowledge, they successfully predicted the PK characteristics of eight small-molecule drugs in rats and monkeys, as well as the PK characteristics of 22 monoclonal antibodies in humans. This approach can also provide crucial mechanistic information during the early stages of drug development, such as predicting non-specific binding of antibody drugs and FcRn receptor interactions, thereby reducing the need for animal testing.

CNS Drug (Central Nervous System) Distribution Prediction:A PBPK model was developed for an Alzheimer's disease antibody drug (tau protein target), accurately predicting the drug’s concentration in rat and monkey cerebrospinal fluid (CSF) and brain interstitial fluid.

Improving Drug Metabolism Prediction Performance by Combining Organoids:Liver-on-a-chip combined with PBPK model successfully predicted the clearance rate of propranolol in humans and accurately captured interpersonal differences in the metabolism of six small-molecule drugs.

Clinical Translation Prediction of Bispecific Antibody (BiTE) Drugs:Researchers successfully predicted the clinical data performance of multiple drugs by constructing a BiTE drug QSP platform model and integrating a large amount of cellular experimental data.

ADC Drug Dose Prediction and Tumor Target Saturation Analysis:Through the Thiele modulus analysis method, the research team clarified the relationship between drug permeability, target density, and drug-target affinity, successfully predicting the target saturation and tolerance of marketed ADC drugs. This method performs exceptionally well in analyzing the clinical tolerance and target saturation of drugs such as antibody-drug conjugates (ADCs) and immune checkpoint inhibitors.

Optimization of Anti-Angiogenic Drug Dosage Selection and Medication Strategies:Digital twin models of the tumor microenvironment have been successfully used to predict the preclinical efficacy of anti-angiogenic drugs, including optimizing dose selection and combination therapy strategies.

Hepatotoxicity Risk Prediction (DILIsym Platform):This platform combines in vitro liver chip experimental data with QST models to successfully predict the hepatotoxicity risks of various drugs and has been widely applied in the FDA drug registration and approval process.

Prediction of Nephrotoxicity Risk:Researchers combined renal chip experimental data with QSP models to accurately predict the clinical renal toxicity profiles of commonly used drugs such as cisplatin and rifampicin.

Cardiotoxicity Prediction:Researchers combined heart chip data with QST models to accurately predict the clinical cardiac toxicity risks of drugs such as doxorubicin and terfenadine.

Hematotoxicity Prediction:Myelosuppression is one of the most common adverse reactions to anticancer drugs, and the QSP model for bone marrow and granulopoiesis has successfully achieved the translation of drug-induced myelosuppression effects from in vitro to humans.

Deep Learning Platform Deep-PK:Deep-PK, a QSAR-based graph neural network platform, successfully predicted pharmacokinetic and toxicity profiles from drug molecular structures, significantly reducing the time cycle for early drug screening.

Digital Twin Model:Animal Digital Twin Model Built on Generative Adversarial Network (GAN) Technology Successfully Predicts Drug-Induced Liver Toxicity Risk, Showing High Consistency with Real Clinical Data.

Acceleration of European and American Policies:In 2022, the United States passed the "FDA Modernization Act 2.0," while Europe issued relevant directives as early as 2010, explicitly promoting the "3R Principles" (Replacement, Reduction, and Refinement of animal testing). These regulations encourage the adoption of new technologies such as cell-based assays or computational models to replace animal testing, thereby obtaining exemptions from animal experiments when studying drug safety and efficacy.

Specific Measures Released:In April 2025, the FDA released a strategic roadmap, stating that it will gradually phase out animal testing in drug development, prioritizing the promotion of NAMs in drugs with clear mechanisms and easy modeling, such as monoclonal antibodies; EMA, NIH, and others have also launched multiple dedicated programs (e.g., Complement-ARIE) to promote the development and standardization of NAMs.

Industry Impact:With regulatory support, NAMs have moved from laboratory exploration to practical applications in the new drug registration and approval process.

In vitro experiment

Conduct preliminary in vitro experiments to obtain key drug parameters such as biochemical characteristics, molecular interactions, and cellular responses;

Model Analysis

Integrate in vitro data into computational models such as PBPK, QSP, and QST, and use these models to analyze and predict species-, drug-, and dose-specific responses;

Model Prediction

Before actual animal testing, computer simulation models are used to predict preclinical pharmacokinetics, toxicity, and biological efficacy;

Small-scale preclinical in vivo experiments

Optimize experimental conditions, dosage selection, and research parameters, and design smaller-scale preclinical studies based on prior predictions;

Continuous Iteration

Iterative feedback as needed, using the results of preclinical in vivo studies to improve in vitro, in silico, and in vivo experiments for enhanced predictive accuracy.