For more than a century, animal experiments have been central to life sciences, facilitating foundational discoveries such as the germ theory of disease and advancing our understanding of conditions like tuberculosis, polio, and rabies. Today, most modern drugs, including penicillin, cancer drugs, and HIV (human immunodeficiency virus) treatments, have first been validated in mice before entering clinical trials and eventually saving lives.

However, while animal studies have yielded valuable insights, their relevance to human disease is limited by profound species differences. These differences mean that as many as 92 percent of drugs that appear promising preclinically ultimately fail in human trials. Many of these, approximately 75 percent, fail due to a lack of efficacy or safety, showing just how poorly animal data often translates to real patients.

Researchers are increasingly adopting human-first approaches that more accurately model human biology. Techniques such as organoids, human induced pluripotent stem cells (iPSCs), and engineered tissues or microphysiological systems, such as organ-on-a-chip platforms, allow scientists to study tissue development, disease mechanisms, and drug responses in a human-specific context. This approach generates insights that are more likely to translate from the bench to clinical outcomes, accelerating the development of safe and effective therapies.

Reflecting this scientific momentum, regulatory frameworks are beginning to shift as well. In late 2022, President Biden signed the FDA Modernization Act 2.0 into law, removing the legal requirement that drugs be tested in animals before entering clinical trials. And in 2025, the agency began phasing out animal testing for monoclonal antibodies and other drugs, encouraging instead the use of human cell-based platforms and simulations using artificial intelligence (AI) and modelling.

Continue reading below...

WebinarsAccelerating rare disease clinical trials

Explore how a rare kidney disease trial achieved faster patient enrollment with data-informed strategies and collaborative partnerships.

Read More

Lorna Ewart, Chief Scientific Officer at Emulate, called this a tipping point. “The FDA Modernization Act 2.0 and the National Institutes of Health’s (NIH) commitment to prioritize funding of New Approach Methodologies (NAMs) over animal-exclusive proposals signal a fundamental shift from ‘permission’ to ‘expectation’ that human-relevant models will play a central role in drug development. Regulators across the globe are not only opening the door to these technologies — they are actively encouraging sponsors to use them.”

To understand why this transition matters, it helps to revisit how mice became so entrenched in the first place.

Why rodents have been the default

For more than a century, mice have been the silent workhorses of biomedical research. Their short lifespans, rapid reproduction, and ease of breeding make them convenient and cost-effective, while their small size and adaptability have allowed labs across the world to house and study them at scale. Crucially, humans and mice also share a very similar genetic background, with around 90 percent of their genomes arranged in conserved regions. Over time, this combination of practical and biological advantages has made the common house mouse the default model for testing everything from vaccines to cancer therapies.

The widespread use of mice has also driven the development of extensive research resources, including thousands of genetically defined inbred strains, a complete reference genome, deep sequencing data for dozens of inbred lines, and detailed maps of genetic variation. Advanced tools for precise genome manipulation allow scientists to replicate human conditions in mice with remarkable fidelity, creating models that are central to preclinical research.

“Efficacy translation is scarcely better than a coin flip across indications, and even safety signals are missed or misprioritized due to species differences, young and healthy inbred strains, artificial disease induction, small cohorts, and short study duration.”
- Thomas Hartung, Center for Alternatives to Animal Testing

However, despite these advantages, mice are not miniature humans. The most obvious difference is size: humans are thousands of times larger than mice, and this affects metabolism, life expectancy, reproduction, and diet. There are also differences in circadian rhythm, cognitive development, sensory systems, and social behavior. Physiological differences extend to organs and tissues as well. Mice differ from humans in relative organ size, structure, regenerative capacity, disease susceptibility, cell duplication time, immune response, and drug transport.

Moreover, the domestication and breeding of laboratory mouse strains have increased these inter-species differences. Inbred laboratory strains lack the genetic diversity seen in human populations, masking variability in drug responses and disease progression. Environmental exposures, diet, and lifestyle, all of which are crucial contributors to human disease, are also absent in standard mouse models.

As Thomas Hartung, Director of the Center for Alternatives to Animal Testing at Johns Hopkins University, told DDN, “Efficacy translation is scarcely better than a coin flip across indications, and even safety signals are missed or misprioritized due to species differences, young and healthy inbred strains, artificial disease induction, small cohorts, and short study duration. This contributes to about 90 percent clinical attrition and late safety surprises.”

These disparities can have profound effects on disease modeling. For example, many anticancer therapies show dramatic success in mice but researchers fail to replicate these results in humans. Given these differences, it is perhaps unsurprising that disease patterns and drug responses in mice often do not reflect what happens in people.

Joseph Wu, Director of the Stanford Cardiovascular Institute and co-founder of Greenstone Biosciences, emphasized that these limitations are driving a shift toward human-relevant models. “Mice have been invaluable for uncovering basic biology, but their predictive power for human drug responses is limited. Cardiovascular physiology differs across species; mice have heart rates five to 10 times faster than humans, distinct ion channel properties, and different metabolism.”

“Mice remain useful for systemic interactions and whole-body physiology,” he said. “But human iPSCs, organoids, and engineered tissues allow us to capture patient-specific genetics and disease phenotypes in ways that mice simply cannot.”

Ewart echoed this sentiment, “Human-based systems such as iPSCs, organoids, and organ chips not only address the species gap by providing data that is directly human-relevant, but they also deliver results faster and at lower cost. This combination of accuracy, speed, and efficiency is what makes them so powerful as alternatives to traditional animal testing.”

The rise of human-based models

One of the most widely used human-based models is iPSCs. First developed in 2006, iPSCs are generated by reprogramming adult somatic cells, such as skin or blood cells, back into a pluripotent state, giving them the ability to differentiate into virtually any cell type. This technology allows researchers to model human tissues and diseases in a way that more reliably mirrors human physiology, disease characteristics, and pharmacogenomics.

At Stanford University, Wu’s lab has built one of the world’s largest iPSC biobanks, now housing over 2,500 lines derived from individuals spanning diverse ages, sexes, and genetic backgrounds. “Traditional mouse models will not recapitulate this genetic variability,” Wu explained. “With large representation of patient-derived iPSCs, we can capture that variability directly in the lab, which helps explain why one patient may respond well to a drug while another develops toxicity.”

iPSC-derived models have opened the door to what Wu calls “clinical trials in a dish.” By testing hundreds of patient-specific cell lines in parallel, his team can identify responders and non-responders before moving into human trials, uncover genetic drivers of drug response, and better predict adverse effects.

Beyond disease modeling, iPSC platforms can be paired with advanced computational tools. Wu highlighted the ADMET-AI platform which “leverages large clinical toxicology datasets. Unlike animal models, it can rapidly forecast human-specific cardiotoxicity risks across thousands of compounds, something preclinical testing in mice simply cannot capture at that scale or with patient relevance. “

Rethinking the role of mice in research

Cardiac fibrosis is a scarring of the heart muscle due to excessive collagen deposition, which impedes heart function, causes symptoms like shortness of breath and swelling, and can lead to heart failure. Unfortunately, no therapies have been approved for this condition. Most screenings for antifibrotic therapies have relied on mouse fibroblasts; however, these models are limited by considerable species differences, the lack of counter-screening for cardiotoxicity, and the inability to replicate contractile function.

Instead, Wu’s team turned to human iPSC-derived cardiac fibroblasts (CFs), patient-specific cells capable of faithfully modeling human disease in vitro. Using a high-throughput screening platform, the researchers tested roughly 5,000 compounds across multiple doses and independent iPSC lines, while counter-screening in iPSC-derived cardiomyocytes and endothelial cells to exclude compounds with potential cardiotoxicity.

Artesunate, a powerful antimalarial drug, emerged as the top hit. In human fibroblasts, it reduced proliferation, migration, and contraction, and decreased collagen deposition. To strengthen translational confidence, the team employed a layered, multiscale approach: engineered heart tissues recreated the 3D cellular and mechanical environment, revealing how fibroblasts and cardiomyocytes interact; mice provided systemic insights into drug metabolism, distribution, and organ-level effects; and computational simulations allowed target identification and drug–protein docking, pinpointing artesunate’s interaction with MD2 (myeloid differentiation factor 2).

This multiscale approach is “absolutely critical because no single model can capture the full complexity of human disease. This multiscale validation gave us the confidence to advance artesunate into early-phase clinical trials for pulmonary arterial hypertension and idiopathic pulmonary fibrosis. More broadly, this layered strategy reflects what regulators are now encouraging under the FDA Modernization Act 2.0 and new NIH NAMs guidance, shifting from reliance on single animal models toward integrated, human-first approaches that combine cellular, tissue, animal, and computational evidence. I believe it’ll become a cost-effective way to reduce late-stage failures and deliver therapies that are truly translatable to patients,” said Wu.

For Wu, these cases highlight a new balance in biomedical research. “Human models give us precision, animal models provide context, and together they make drug development more efficient,” he says. Mice still play an important role, but increasingly as part of the supporting cast in a research pipeline that now begins with human-relevant models.

From cells to miniature organs

Advances in stem cell culture have made it possible to generate large, uniform populations of specific human cells. These can be assembled into engineered tissues or organ-on-a-chip systems, enabling scientists to study organ-level physiology and disease in ways that were once impossible outside the human body.

“Mouse models have taught us a lot about organ formation,” Wu notes, “but there are limits to what we can learn about human development in vivo. Ethical restrictions and species differences mean that the earliest stages of organogenesis remain largely uncharted in humans.”

Using spatially micropatterned iPSCs and four fluorescent reporter systems, Wu’s lab has guided cells to self-organize into cardiac and hepatic organoids with lumenized, branched vasculature. These organoids, or “gastruloids,” mimic key aspects of human embryonic development during the first three weeks post-conception, revealing how the heart and liver begin to form their intricate structures.

Current gastruloids lack full maturation, long-range vascular perfusion, functional innervation, and the dynamic immune-stromal interactions that shape organ development and disease. The next frontier is to integrate hierarchical vasculature, neural input, and multicellular niches to move them closer to physiological fidelity.
– Joseph Wu, Stanford Cardiovascular Institute

A carefully optimized cocktail of growth factors and small molecules guided the formation of branched, lumenized vasculature within organoids containing multiple interacting cell types, including endocardial, myocardial, epicardial, and neuronal cells. These organoids reproduce key structural and functional features of early human organs, allowing researchers to study vascularization, cell-cell interactions, and developmental signaling in a human-specific context.

However, these models are not perfect. “Current gastruloids lack full maturation, long-range vascular perfusion, functional innervation, and the dynamic immune-stromal interactions that shape organ development and disease. The next frontier is to integrate hierarchical vasculature, neural input, and multicellular niches to move them closer to physiological fidelity,” noted Wu.

Still, they offer a human-specific lens that no mouse model can provide, allowing researchers to study early development, model diseases, and test drugs in ways that were previously impossible. “The challenge now is systematically closing these gaps so that they can serve not only as discovery platforms but also as credible tools in the drug development pipeline,” said Wu.

Limits of current human models

Even as iPSCs, organoids, and gastruloids reshape drug discovery, the path is far from straightforward. Hartung cautioned that scientific and practical hurdles remain: “We still need broader qualification of human models, quantitative links to clinical endpoints, and consensus pathway mapping for toxicities. On the practical side, standardization, cost-of-entry, workforce training, and regulatory familiarity remain uneven.”

One major hurdle is scalability. “Regulators and sponsors need large, reproducible datasets before they can replace entrenched animal models. Until recently, the throughput of NAM technologies was simply too low to generate that level of evidence. But with new platforms capable of supporting dozens to hundreds of replicates per run, that barrier is coming down,” said Ewart.

Generating patient-specific iPSCs, differentiating them into specialized cell types, and maintaining large biobanks demands time, space, and resources. While pooling hundreds of lines can increase throughput, differing growth rates and functional readouts can complicate results, sometimes masking patient-specific responses. Recent advances in single-cell technologies are helping to overcome this. By barcoding each line through whole-genome sequencing, researchers can track individual cells in pooled experiments and link gene expression data back to the original donor. This approach allows high-throughput analysis across diverse genetic backgrounds, bringing large-scale preclinical testing in human iPSCs closer to reality.

Reproducibility is another challenge. Differentiation protocols can yield subtle variations across lines or labs, affecting disease modeling and drug testing outcomes. Standardization is improving, but large-scale, multi-site studies are still limited by variability in cell quality and maturation.

Finally, regulatory integration remains a moving target. The FDA Modernization Act 2.0 and NIH NAMs guidance support human-first evidence, but translating that data into clinical approvals still demands strategic validation, likely using a combination of iPSCs with engineered tissues, computational models, and selective animal studies.

So, can we stop using mice for research? The answer may be yes — but not completely, at least not yet. Instead, their role is evolving from central to supportive, as human-first platforms take the lead in predicting drug responses, understanding disease, and ultimately bringing safer, more effective therapies to patients.