Artificial intelligence can detect toxicity of chemicals

Swedish scientists have developed an artificial intelligence (AI) method to identify toxic chemicals based solely on knowledge of their molecular structure. The technology’s application will help understand and control an increasing number of chemicals used, many of which are toxic, thereby reducing the need for animal testing.

From household products to industry, chemicals are ubiquitous. As circulation is a natural process, they find their way into water and ecosystems, where they can cause serious harm to the environment and human health.

Despite numerous regulations and animal testing to determine chemical safety, adverse effects persist. In the European Union alone, over two million animals are used annually to comply with various chemical regulations.

Swedish researchers from Chalmers University of Technology (CTH) and the University of Gothenburg have developed an AI method that enhances the identification of toxic chemicals based solely on knowledge of the molecular structure. They pointed out that the application of artificial intelligence is fast and cost-effective. The study’s authors claim it can determine toxicity in the early stages of chemical development, thereby reducing the need for animal testing.

There are more than 100,000 chemicals on the market

“Our method can predict whether a substance is toxic or not based on its chemical structure. It has been developed and refined by analyzing large datasets from laboratory tests performed in the past. The method has been trained to make accurate assessments for previously untested chemicals,” researcher Mikael Gustavsson says.

According to Professor Erik Kristiansson, there are currently over 100,000 chemicals on the market, but only a small part have a well-described toxicity towards humans or the environment.

He emphasized, “Assessing the toxicity of all these chemicals using conventional methods, including animal testing, is not practically possible. Here, we see that our method can offer a new alternative.”

Scientists believe the method can be highly useful for environmental impact assessments, as well as for authorities and companies using or developing new chemicals. Thus, their AI program is open and available to the public through the Trident service.

ChatGPT for chemicals

Computational tools for finding toxic chemicals exist, but their applicability and accuracy are limited. In their study ‘Transformers enable accurate prediction of acute and chronic chemical toxicity in aquatic organisms’, published in Science Advances, researchers compared the AI method they developed with three of the most commonly used computational tools. The results show that the AI method is more accurate and applicable.

“Our results show that AI-based methods are already on par with conventional computational approaches, and as the amount of available data continues to increase, we expect AI methods to improve further,” Kristiansson said.

Artificial intelligence systems will be able to largely replace laboratory tests

Researchers estimate that artificial intelligence systems would be able to largely replace laboratory tests, reducing the number of animal tests as well as testing costs.

As the study’s name suggests, the method is based on transformers. It is a deep-learning AI model originally developed for language processing. ChatGPT is an example of such an application.

The model was also proven highly effective in extracting information from chemical structures. Transformers can better identify properties in molecular structures that cause toxicity than was previously possible.