Next-Gen Preclinical Translation: NAMs, Humanized Mouse Models, and AI Integration Take Center Stage at BIO Asia-Taiwan 2026

The preclinical drug discovery session at BIO Asia-Taiwan 2026 convened toxicologists, translational scientists, and regulatory experts to evaluate a structural transition reshaping pharmaceutical safety profiling. Hsian-Jean Chin, Director General of the National Center for Biomodels at NIAR, framed the discussion by noting that drug discovery has always been a race against cost and time. Chin underscored a stark reality: nearly 90 percent of therapeutic candidates fail in human clinical trials due to lack of efficacy or unforeseen toxicity, despite passing traditional animal testing.

Traditional animal models frequently fail to recapitulate complex human organ biology, human-specific receptors, or intricate immune responses, imposing massive financial losses and clinical delays on global biopharmaceutical companies. To address this translational gap, the conference highlighted the rapid adoption of New Approach Methodologies (NAMs), including humanized mouse platforms, organ-on-a-chip technologies, and artificial intelligence algorithms. Legislative initiatives such as the US FDA Modernization Act 2.0 now provide formal regulatory authorization to utilize validated non-animal methods for Investigational New Drug (IND) applications.

Humanized Mouse Models Bridge the Translational Gap

Hsian-Jean Chin presented breakthroughs in engineering humanized mouse platforms for oncology and autoimmune drug testing. Traditional immunodeficient mice lack functional immune cells, rendering them incapable of evaluating immuno-oncology agents like checkpoint inhibitors, bispecific antibodies, and CAR-T therapies. Engrafting human CD34 (搜索)搜索该靶点+ hematopoietic stem cells into severely immunodeficient backgrounds yields models with functional T-cells, B-cells, natural killer cells, and myeloid lineages.

Chin demonstrated how these models accurately predict human-specific immune responses, including cytokine release syndrome (搜索)搜索该疾病 (CRS) and immune cell-mediated organ toxicities that go undetected in standard rodent models. He also highlighted advanced patient-derived xenograft (PDX) humanized platforms that co-engraft patient tumor tissue with matching human immune cells, enabling drug developers to evaluate personalized immuno-oncology responses and immune evasion mechanisms in vivo.

Brandy Wilkinson, CEO of GemPharmatech (搜索)搜索该公司, reinforced that researchers will continue using mice because purely in vitro systems cannot yet fully replace living organisms. She noted that recent advances in gene therapy and PROTACs mean virtually every biological target is now accessible, requiring sophisticated in vivo platforms. Wilkinson urged the scientific community to rigorously re-evaluate and re-validate standard mouse strains to eliminate laboratory inconsistency.

Organ-on-a-Chip Technologies Demonstrate High Predictive Accuracy

Dr. Donald Ingber, Founding Director of the Wyss Institute at Harvard University, presented advancements in microphysiological systems (MPS) and organ-on-a-chip technologies. These microfluidic chips, lined with living human parenchymal and vascular endothelial cells subjected to physiological fluid flow and mechanical stretch, recreate functional human tissue microenvironments. The bioengineered chips accurately mimic organ-level physiology, tissue-tissue interfaces, and hemodynamic forces characteristic of vital human organs, including the liver, lung, kidney, intestine, and blood-brain barrier.

In landmark validation studies, human liver-on-a-chip models accurately identified toxic drugs that had passed traditional animal safety screens with false-negative results, achieving over 87 percent sensitivity and 100 percent specificity for human hepatotoxicity. Dr. Ingber emphasized that connecting multiple organ chips via circulating vascular channels establishes multi-organ human body-on-a-chip systems, allowing researchers to evaluate complex organ-organ toxicological interactions and systemic drug metabolism prior to human testing.

Matt Hewitt, Vice President and CTO of Manufacturing Business Division at Charles River, noted that while organ-on-a-chip platforms excel at isolated target validation, predicting multi-organ toxicities across complex organ systems remains challenging. Charles River is advancing Next-Generation Sequencing (NGS) as a non-animal method for viral safety and cell characterization, reducing testing turnaround times from twelve weeks to five weeks while detecting one infected cell in a million.

AI-Driven Predictive Toxicology Accelerates Screening

Yun Yen, Chair Professor at Taipei Medical University, highlighted that AI algorithms can now screen libraries of billions of compounds against biological targets in just two days—a process that previously required a full year. Machine learning algorithms analyze chemical structures to identify specific toxicophores and predict off-target interactions, allowing medicinal chemists to filter out virtual compounds with high predicted cardiotoxicity, mutagenicity, or hepatotoxicity before physical synthesis.

Integrating AI models with high-throughput transcriptomic profiling data from human organ-on-a-chip assays can reduce preclinical lead optimization timelines by up to 50 percent. However, Yen offered a crucial reality check, citing the clinical trial failure of Exscientia's AI-designed candidate, which entered clinical trials in twelve months but failed due to unforeseen human pharmacokinetics. "While AI accelerates compound selection, computational speed does not guarantee clinical efficacy," Yen emphasized, reinforcing that AI is an augmenting tool requiring human scientific expertise.

Tsai-Kun Li, President of the Development Center for Biotechnology (DCB), outlined Taiwan's strategic initiative to combine AI algorithms with national health data, aiming to transform the regional biotechnology ecosystem. DCB integrates AI across target identification, nucleic acid design, and neoantigen discovery to accelerate early-stage drug development and de-risk academic discoveries up to the IND stage.

Market Growth Reflects Industry Momentum

The AI-accelerated nonclinical drug testing market is valued at USD 1,090.7 million in 2026 and is forecast to reach USD 4,489.9 million by 2036, growing at a 15.2% CAGR. Machine learning platforms account for an estimated 30.5% share in 2026, while toxicity prediction represents 31.4% of application revenue. In silico testing is projected to capture 39.0% of the test type segment. The United States is projected to record 16.3% CAGR through 2036, while South Korea is predicted to advance at 17.0% CAGR, reinforced by public programs supporting AI skills across drug development teams.

Regulatory Frameworks and the Path Forward

Regulatory experts addressed qualification frameworks necessary for formal acceptance of non-animal methods. While agencies support NAM adoption, developers must provide rigorous analytical validation proving novel in vitro and in silico assays are reproducible, robust, and predictive of human clinical outcomes. International initiatives, including the FDA's Predictive Toxicology Roadmap and the European Medicines Agency's Innovation Task Force, are establishing standardized qualification programs.

Matt Hewitt cautioned that regulators demand complete transparency rather than black box models, reminding scientists that human researchers remain fully accountable for clinical safety because algorithms cannot bear legal or moral responsibility. Panelists concluded that establishing collaborative consortia between pharmaceutical companies, academic bioengineers, and regulatory authorities to generate standardized validation datasets is essential for widespread industry adoption. Replacing traditional animal testing with validated human-relevant non-animal methods will dramatically reduce drug development costs, accelerate clinical entry, and improve clinical trial success rates for innovative therapeutics worldwide.