Can Artificial Intelligence Reduce Traditional Animal Testing?
Image created by author with Image Creator by Microsoft Designer
This article addresses the remarkable potential of artificial intelligence (AI) to reduce traditional animal testing (AT). Not all lab animals will be as lucky as our mouse friend pictured above, but with AI, more and more will.
Traditional AT
For decades, AT has persisted in many areas, including toxicity testing, drug discovery, and disease-related studies. Government regulations have often required AT and some private companies and research institutes have voluntarily performed it.
Problems with AT
Four main problems haunt AT:
Cruelty: AT often causes death or severe harm and suffering to lab animals. We’ve all seen the photos, heard the distress, noticed the fear, and felt sickened. Many billions of animals have died in testing.
Ineffectiveness: AT regularly fails to accurately predict outcomes in human clinical trials due to differences between humans and lab animals, artificial laboratory conditions, dosing challenges, limited ability to model complex human diseases, and animal stress factors. Approximately 90 percent of drugs passing safety and efficacy tests in AT fail during human trials.
Time-Consuming: AT is often very slow. For example, drug-related AT generally takes several years, meaning delays in drugs needed by patients.
High Costs: AT is generally more expensive than AI, sometimes costing millions. One way or another, consumers bear these costs.
AI's emergence as an alternative to AT
Independently and integrated with other technologies, AI has emerged as a potentially game-changing alternative to AT. Several forces have propelled its advance.
AT Problems: As outlined above, the significant issues with AT have driven the search for alternatives.
AI Advancements: The exponential growth in AI capabilities, fueled by algorithmic advancements, computing power, and data availability, has made AI a viable alternative.
Legislative and Regulatory Embrace: Recent legislative and regulatory actions have supported AI as an AT substitute. Here are some examples:
The federal FDA Modernization Act 2.0 became law in late 2022, allowing drugmakers to substitute scientifically rigorous alternatives for AT. While optional and the FDA may still reject alternatives as insufficient, this law represents significant progress.
The EPA, historically a major user of AT for toxicity testing, announced in 2019 it would "aggressively reduce animal testing" to eliminate all routine safety tests on mammals by 2035.
The federal Interagency Coordinating Committee on the Validation of Alternative Methods (ICCVAM), with 17 key federal regulatory and research agency members, has intensified its efforts to use AI as an AT substitute.
The FDA recently created the AnimalGAN Initiative, a program designed to advance AI-based virtual animal models as AT alternatives.
AI alternatives and benefits
AI-based or integrated alternatives to AT encompass a fascinating array with futuristic and evocative names like organ-on-a-chip, digital twins, 3D printed organs, patient-derived cell models, high-throughput screening, silicon modeling, digital simulations, virtual clinical trials, and microfluidic devices. For scientists and lab animals alike, these technologies are fostering an evolving jailbreak from the prison of AT tradition. We can’t cover all of those technologies in this article, but here are a few examples where AT is most commonly used: toxicity, drug development, and disease-related matters.
Toxicity Testing: Toxicity testing aims to assess how much a substance can harm a human or other organism. Many items undergo toxicity testing, including cosmetics, pesticides, food additives and drugs. AI-based computational models — a way to study complex systems through the power of computers — offer a promising alternative to AT. They use machine learning algorithms that learn from vast databases of existing toxicological data, including chemical structures, biological interactions and historical testing results, to predict the toxicity of new compounds. These algorithms can operate at a fraction of AT’s time and often expense. Moreover, they're constantly fed new data, increasing their odds of more accurately evaluating toxicity than AT.
Drug Discovery: AI is revolutionizing drug testing by leveraging computational algorithms to analyze molecular structures, predict drug-target interactions, and identify potential side effects. AI also enables the development of personalized medicine approaches — that is, treatments tailored to individual patients based on their genetic makeup, disease characteristics, and lifestyle factors. Because AI analyzes data and predicts outcomes much quicker than traditional methods, it accelerates drug development timelines, potentially saving lives and improving patient outcomes. It also eliminates AT costs, such as animal care and housing, making drug development more cost-effective and potentially leading to more affordable medications.
Disease-Related Studies: AI also offers new avenues for studying diseases that can significantly reduce AT. One is disease modeling — AI algorithms that analyze vast datasets of medical records, genomic data, and molecular interactions to create virtual models of human diseases. These models may offer a more accurate understanding of disease progression than AT, which often fails to adequately model the complexities of human disease due to species-specific differences in biology and physiology. AI-powered virtual models, on the other hand, can be tailored to specific human conditions, leading to more relevant and targeted research strategies. AI can also analyze an individual's medical history, lifestyle factors and genetics to identify disease risks, predict individual medication responses, and tailor treatment plans accordingly.
Challenges and Future Considerations:
While AI holds immense promise for reducing AT, it faces the following challenges.
Data Accuracy: The effectiveness of many AI technologies depends on the quality and quantity of data used for training. This requires robust data collection and curation practices
Model Interpretability/Transparency: Understanding the reasoning behind AI predictions is critical for building trust and ensuring ethical implementation. Advancements continue in explainable AI — AI that attempts to explain how it works, but the black box nature of AI continues to spook some.
Custom: Persuading stalwart supporters of AT to transition to non-AT methods will be an uphill battle.
Regulatory Acceptance: Ensuring reliable and valid AI predictions and fostering wider adoption also hinges on regulatory bodies establishing clear guidelines and frameworks for incorporating AI-powered methods into safety assessments and drug development processes.
Conclusion
Yogi Berra memorably cautioned that “prediction is very difficult, especially about the future.” It’s hard to argue with that when considering the ultimate impact of AI on AT. Yet, given the problems with AT and the capabilities of AI, it seems highly likely that AI will be good news for many lab animals and, in turn, for the health of many humans and other animals.
This article is based on many scholarly papers, news articles and other information, including the recommended articles below.
Next month, we’ll investigate the potential of AI to bridge the communication gap between humans and non-human animals.
Frank Brown, UU of Arlington, Virginia
Further Reading:
Artificial Intelligence (AI)-it’s the end of the tox as we know it (and I feel fine)
The Johns Hopkins Center for Alternatives to Animal Testing
Meet the AI animal testing alternative that could prevent unnecessary deaths
Alternatives to Animal Testing
Replacing Animal Testing: How and When?