Artificial intelligence: a tool able to evaluate the toxicity of chemical substances

Inserm researchers have developed a system capable of identifying the toxic effects of bisphenol S. The chemical compound increases the risk of obesity.

Artificial intelligence is the ally of researchers. A study conducted by scientists from Inserm and published in Environmental Health Perspectives confirms it. The team, led by Karine Audouze, has developed a computer tool capable of identifying the toxic effects of bisphenol S. This substitute for bisphenol A, now banned, promotes obesity.

New on the side of # IA that is useful for assessing the toxicity of chemicals. Example with bisphenol S //t.co/cQkD0CcWks pic.twitter.com/Ex3v4IuITM

- Inserm (@Inserm) April 17, 2019

Compilation of scientific data

Karine Audouze and her team from the Inserm unit UMR-S1124 (for Environmental Toxicity, Therapeutic Targets, Cell Signaling and Biomarkers) have developed the AOP-helpFinder computer program. The latter is able to determine the toxicity of the products by compiling the data present in the scientific literature. The tool identifies occurrences of terms in scientific texts: those that designate chemicals and those that designate pathologies. He is then able to analyze the relations between the different terms. "Beyond a quick reading, the system allows a true analysis of automated text," says the scientist in a statement. Depending on the positioning of the term, AOP-helpFinder determines the strength of the link. For example, if both words are at the end of a text, the "weight" attributed by artificial intelligence to the link is more important because it is likely to be the conclusion. Conversely, when the two words are at the beginning of the text, it is more plausible that they are hypotheses, and that the link is weak.

The bisphenol S test

Bisphenol S has been used as a tool for validating the computer program by researchers. They have incorporated in artificial intelligence all the possible names and synonyms of the substance. Then, scientific articles were presented to the software. "The goal," says Karine Audouze, "was to establish links between the terms representing the chemical and those corresponding to the pathological processes." The results indicate a correlation between bisphenol S and the risk of obesity. The researchers manually tested these findings and confirmed them: bisphenol S increases the risk of adipocyte formation, the cells that store fat.

"(The program) does not, as such, evidence toxicity, concludes the scientist, but serves to quickly integrate a large amount of information and prioritize the most likely adverse effects, allowing to design the most relevant biological and epidemiological studies ". The study will now continue, the researchers want to create a second version of the tool able to analyze not only a single substance but a set of compounds.

Multiple uses

The applications of artificial intelligence in the field of health are endless. In January 2019, researchers succeeded in creating a tool that could detect the first symptoms of Alzheimer's disease. In Britain, another research team has managed to use artificial intelligence to predict premature death.

Video: MD vs. Machine: Artificial intelligence in health care (January 2020).