Human Cells Are Rewriting the Rules of Drug Development: iPSCs, Organoids, and AI Converge to Challenge Animal Models

Drug development is approaching a significant inflection point. Advances in induced pluripotent stem cells (iPSCs), organoids, and organs-on-chips are converging with artificial intelligence and a rapidly changing regulatory environment to challenge pharmaceutical R&D's longstanding dependence on animal models, according to Nina Bauer of FUJIFILM Cellular Dynamics (搜索)搜索该词条.

In a recent conversation with BioSpectrum Asia, Bauer outlined how non-animal methods (NAMs) have gained significant momentum, driven by a confluence of scientific validation, technological maturity, and regulatory modernization that is fundamentally reshaping how medicines are discovered, tested, and brought to patients.

The Regulatory Tipping Point

The shift toward NAMs represents a convergence of multiple forces. Bauer pointed to a pivotal FDA-led workshop in 2013 that launched the Comprehensive In Vitro Proarrhythmia Assay (CiPA) initiative. "It proved truly groundbreaking in that it ultimately demonstrated that human iPSC-derived cardiomyocytes, combined with in silico modeling, could predict cardiac safety better than animal models," Bauer said. "Importantly, the regulators accepted it. This proof-of-concept opened the doors to the development of what is now commonly thought of as NAMs."

Regulatory support has accelerated dramatically in recent years. The FDA Modernization Act 2.0, passed in 2022, eliminated the animal testing mandate and explicitly recognized NAMs as valid alternatives. The European Medicines Agency followed with 3Rs implementation strategies and guidance on iPSC-based models, while the UK's MHRA created expedited pathways for NAMs submissions post-Brexit. Most recently, China's NMPA announced it would accept validated NAMs for specific endpoints.

"We are at a significant inflection point, where the underpinning scientific rigor exists, the technology is accessible and scalable, and regulators globally are not just accepting but actively encouraging these human-relevant approaches," Bauer stated.

iPSCs as the Backbone of Next-Generation Models

iPSCs are increasingly viewed as the foundation of human-relevant models because they address three critical limitations of conventional in vitro systems. First, they are created from stable, scalable master cell banks that can be differentiated on demand into many different cell types, enabling the industrialization of assays. Second, they capture human genetics more faithfully, allowing researchers to build panels of iPSC lines reflecting diverse patient genotypes, specific disease mutations, and rare variants. Third, the combination of iPSCs with gene editing tools such as CRISPR has dramatically accelerated model creation.

"Engineering isogenic disease and control lines can be done in weeks to months, instead of the year or more needed to generate and backcross a comparable rat or mouse model," Bauer explained.

The impact is visible across multiple domains. Beyond the CiPA cardiac safety measures, which have been incorporated into International Council for Harmonization guidelines, developers are using iPSC-based CNS disease modeling and neurotoxicity assays, liver metabolism and drug-induced liver injury risk assessments, and increasingly, multi-organ microphysiological systems.

Organoids and Organs-on-Chips Transform Preclinical Research

Human organoids, assembloids, and organs-on-chips are transforming preclinical research by providing 3D, multicellular systems that more closely recapitulate human tissue architecture, function, and microenvironments than flat cell cultures or animal models. These platforms enable recreation of patient-specific and genetically defined pathologies, including tumor microenvironments, neurodevelopmental disorders, and fibrotic lung pathologies.

In safety testing, these systems provide access to more sensitive and mechanistic toxicity assessments. "Human liver, kidney, cardiac, and brain models can capture delayed toxicities and idiosyncratic ones that are unpredictable because they're either patient-specific or rare, all of which are often missed in animals," Bauer noted.

The fastest adoption is occurring in oncology, where tumor organoids and immune co-cultures are being deployed; neurology and psychiatry, leveraging brain organoids and assembloids; cardiometabolic disease, using heart, liver, and vascular chips; and respiratory and GI indications, where organoid biobanks and standardized chips are being integrated into pharma pipelines and early regulatory dialogues.

AI and iPSCs: A Complementary Convergence

Bauer emphasized that AI and iPSC-derived models are highly complementary, each solving a critical weakness of the other. AI is powerful but fundamentally limited by the lack of extensive, comprehensive, high-quality datasets — the "garbage in, garbage out" problem. Most historical drug discovery data is sparse, heterogeneous, and biased toward non-human systems, with almost no negative data available.

"iPSCs, by contrast, can now provide a stable, scalable, and genetically well-defined source of human cells across tissues and patient genotypes, generating cleaner, more comprehensive, and longitudinal datasets," Bauer said.

Pairing standardized iPSC models with high-content readouts and multimodal omics can create the rich training data AI requires, potentially enabling digital twins and moving early discovery into an in silico environment. "Iterative design-test-learn cycles, which today take a long time and are a key limiting financial factor in drug development, can be moved to an AI model to propose candidates or mechanisms that can then be quickly validated in iPSC-based systems, thus closing the loop between computation and biology."

Remaining Hurdles

Despite rapid progress, significant challenges remain. Scientifically, many models still lack full tissue complexity, including vascularization, immune components, stromal interactions, and the ability to mimic chronic exposure. Manufacturing iPSC-based NAMs at scale with GLP compliance, tight QC, batch-to-batch consistency, and standard protocols remains a work in progress.

Validation represents a central bottleneck. "We need robust, specific evidence to demonstrate that NAMs can be as predictive, reliable, and reproducible as the existing animal models we are intending to replace," Bauer said. She called for standardized protocols, multi-site trials, well-designed concordance studies against historical animal and clinical data, and transparent performance metrics.

Global regulatory harmonization also remains incomplete. "ICH/OECD-style frameworks between the different regulatory agencies might help with streamlining development and data acceptance. Without this level of rigorous, shared validation, NAMs will remain attractive supplements rather than true replacements, especially for high-stakes decisions like first-in-human-enabling toxicology and release testing."

The Five-Year Outlook

Looking ahead, Bauer predicted three main trends: increased maturity and standardization of iPSC-based NAMs with GLP-aligned, ready-to-use assay kits; greater automation in cell manufacturing and NAMs application enabled by AI; and gradual replacement of animal-based testing as regulatory comfort grows.

To remain competitive, Bauer advised biopharma companies to build strong partnerships with trusted suppliers, develop internal assay capabilities on industrialized workflows using robotics and automation, invest in analytical and bioinformatics capabilities, and engage proactively with regulators around specific iPSC use cases.