11K Views 2 years ago

IBM Researchers Want to Test AI Instead of Lab Animals

A team of scientists, including researchers from IBM, has introduced a new artificial intelligence model aimed at transforming the way drugs and substances are tested. The innovative model replaces the use of living animals in toxicity tests, addressing long-standing ethical concerns and scientific limitations associated with animal testing.

Source: HSUS/YouTube

Traditional toxicity testing on animals, which often involves force-feeding, vomiting, paralysis, convulsions, and internal bleeding, has been a contentious issue for years. These tests are not only ethically troubling but also scientifically inadequate, as the responses of animals do not always correlate with those of humans. One infamous example is the LD50 test, which required scientists to feed animals substances until half of them died, without any pain relief. Millions of animals are subjected to such experiments in the European Union and the United States annually.

Over the years, researchers have explored alternatives to animal testing, including organ-on-a-chip technology and other AI-based models. The latest AI model, developed by a team of researchers in the United States and India, shows promise in both improving the reliability of toxicity testing and eliminating the need for animal subjects.

The AI model was trained using data from approximately 50,000 molecules, allowing it to distinguish between toxic and non-toxic structures. What sets this model apart is its reliance on “pertinent positives” and “pertinent negatives,” which helps predict how toxic a substance might be for humans. Unlike animal tests that rely on how animals react to a substance and then extrapolate the results to humans, this AI model directly analyzes molecular properties and their potential impact on human health.

A distinctive feature of this AI model is that it solely utilizes historical data from previous animal experiments, eliminating the need for new animal testing. The model can accurately predict human clinical toxicity without inflicting harm on animals. In contrast, other AI initiatives focused on creating virtual animal models may require ongoing animal test data.

While the scientific community recognizes the limitations of animal data, many still consider it the gold standard for predicting human toxicity. Regulatory agencies have made progress in reducing the demand for animal testing, but there is a lack of clear guidance on how non-animal testing alternatives can satisfy data requirements. This can lead to delays in getting products to market.

To overcome these challenges, regulatory bodies must undergo a full paradigm shift. They should acknowledge that newer non-animal tests cannot be one-for-one replacements for older animal tests. Instead, a range of assessment tools that are more specific and accurate for humans should be embraced. This transition will require extensive training to shift the focus from assessing adverse effects in animals to understanding the modes of action relevant to human safety.

The AI model has garnered interest and promise, building on data from existing animal research, in vitro lab tests, and human clinical data. However, some experts believe that regulators may still require some animal data before authorizing human trials, as a measure of safety for human volunteers.

Sign this petition to stop animal testing for cosmetics!