This Computer Program Could Make Animal Testing Obsolete

This story is part of Treehugger's news archive. Learn more about our news archiving process or read our latest news.
CC BY 2.0. GrrlScientist

Using artificial intelligence, it's now possible to map out previously unknown relationships between molecular structure and chemical toxicity.

A new computer system has been developed in the United States that predicts the toxicity of chemicals more accurately than animal tests. It is a breakthrough development that could potentially reduce the need for tests that are considered highly unethical by many, as well as being expensive, time-consuming, and often inaccurate. As I wrote earlier this year,

"An estimated 500,000 mice, rats, guinea pigs, and rabbits are used each year for cosmetics testing. Tests include assessing irritation, by rubbing chemicals into animals' eyes and skin; measuring toxicity, by force-feeding chemicals to animals to determine if they cause cancer or other diseases; and lethal dose tests, which determine how much of a substance is needed to kill an animal."

The computer-based system offers an alternative approach. Called Read-Across-based Structure Activity Relationship, or "Rasar" for short, it uses artificial intelligence to analyze a database on chemical safety that contains the results of 800,000 tests on 10,000 different chemicals.

The Financial Times reported,

"The computer mapped out previously unknown relationships between molecular structure and specific types of toxicity, such as the effect on the eyes, skin or DNA."

Rasar achieved 87 percent accuracy in predicting chemical toxicity, compared to 81 percent in animal tests. The results were published in the journal Toxicological Sciences, while its lead designer Thomas Hartung, a professor at Johns Hopkins University in Baltimore, presented the findings at the EuroScience Open Forum in France last week.

Companies that produce chemical compounds would eventually be able to access Rasar, which will be made available to the public. When formulating something like a new pesticide, the manufacturer could pull up information about various chemicals without having to test them individually. Duplicative testing is a real problem in the industry, Hartung said:

“A new pesticide, for example, might require 30 separate animal tests, costing the sponsoring company about $20 million... We found that often the same chemical has been tested dozens of times in the same way, such as putting it into rabbits’ eyes to check if it’s irritating."

Some concerns have been raised about criminals being able to access the database and using the information to make toxic compounds of their own, but Hartung thinks there are more direct ways of getting that information than navigating Rasar. And the benefits to the chemical industry (and lab animals) arguably outweigh the risks.

Rasar sounds similar to the Human Toxicology Project Consortium, which I wrote about after attending the Lush Prize in London last fall. HTPC is also working to build a database of information about chemicals, based on results from toxicity and exposure tests and predictive computer programs. This approach is called Pathway-Based Toxicology, and its goal is to make animal testing obsolete while offering better predictions about chemicals' reactions in the human body.