Cage-to-code: From animal experimentation to AI-driven drug discovery

The shift from animal-based to AI-driven drug discovery presents opportunities and challenges in biological interpretation, regulatory acceptance, and education. This review examines the transition toward AI-enabled, human-centric drug discovery within the 3R framework. It explores global regulatory drivers, scientific gaps from reduced animal use, and AI's role in reconstructing biological complexity via multi-omics and predictive modeling. AI-driven polypharmacology offers a scalable, ethical, and precision-oriented framework for future pharmacological research.

Pharmacology has evolved from empirical remedies to systems-level drug discovery. Single-target drugs often fail for complex diseases involving dysregulated networks. Polypharmacology—modulating multiple targets—has emerged as a rational strategy. Ethical and regulatory pressures to reduce animal use, guided by the 3Rs, have accelerated alternatives. However, replacing animal models means losing organism-level complexity. AI and computational biology are bridging this gap, enabling a "Cage-to-Code" transition. Key milestones include high-throughput screening, combinatorial chemistry, QSAR, and polypharmacology. The shift from in vivo to in silico and in vitro methods reflects ethical and scientific pressures. In vitro models are cost-effective but lack systemic complexity, while in silico approaches enable rapid ADMET and polypharmacology prediction.

Over 40 countries have banned cosmetic animal testing. The FDA Modernization Act 2022 mandates New Approach Methodologies. The EU banned cosmetic testing in 2004/2009; India (2014), Brazil (2023), and Canada (2023) followed. OECD increasingly accepts validated alternatives.

Implications of the Ban
Reduced animal use impacts pharmacology training and drug development, creating gaps in practical skills and systemic modelling. Alternatives like organoids, organ-on-chip, and computer modelling are widely used, but no single method fully replicates whole-organism complexity. Incomplete models, educational challenges, regulatory bottlenecks, and limited high-throughput capacity remain significant obstacles.

The Rise of AI as a Solution
AI integrates multi-omics with deep learning to model drug effects. Key applications include predictive modelling, accelerated discovery, personalised medicine, regulatory support, and virtual education. Notable platforms include DeepChem, AlphaFold, BIOVIA, and Pharma.AI. Machine learning works for medium datasets; deep learning (CNNs, RNNs, GNNs) excels at multi-target relationships.

Data leakage overestimates performance; scaffold splitting ensures generalisation. AI platforms like REMEDi4ALL and DeepDrug support drug repurposing; zidovudine repurposing for cancer is a notable example. POLYGON uses generative AI for dual-target design. AI also enables real-time adverse event detection through pharmacovigilance. Future prospects include de novo design using generative AI combined with AlphaFold, clinical trial simulations, multi-omics integration for biomarker discovery, and AI-driven clinical decision support systems.

Limitations and Ethical Challenges
AI relies on sensitive data; algorithmic bias can affect underrepresented groups. "Black box" models lack interpretability. AI cannot fully replace animal testing yet, as animal studies still reveal whole-body responses (long-term toxicity, immune interactions) that AI cannot replicate. AI remains complementary within the 3Rs. Regulatory acceptance requires validated, interpretable outputs; explainable AI and reinforcement learning with multi-objective reward functions are future priorities.

In education, virtual laboratories replace live animal experiments in pharmacology teaching, offering interactive substitutes with instant feedback, though blended learning approaches are recommended.

Conclusions
The shift from animal experiments to AI-driven discovery represents a pivotal evolution. The future lies in an integrated "Cage-to-Code" paradigm where AI orchestrates in vitro, in silico, and limited in vivo validation to satisfy the 3Rs without compromising human safety. As regulatory frameworks evolve, pharmacology must transition to human-centric predictive modelling, fostering a more humane and efficient therapeutic landscape.

Full text

https://www.xiahepublishing.com/2572-5505/JERP-2025-00058

The study was recently published in the Journal of Exploratory Research in Pharmacology.

Journal of Exploratory Research in Pharmacology (JERP) publishes original innovative exploratory research articles, state-of-the-art reviews, editorials, short communications that focus on novel findings and the most recent advances in basic and clinical pharmacology, covering topics from drug research, drug development, clinical trials and application.

Follow us on X: @xiahepublishing

Follow us on LinkedIn: Xia & He Publishing Inc.