Researchers from Texas A&M University have developed a neural network that quickly assesses the toxic potential of chemical compounds. The algorithm analyzes the structure of molecules, comparing them against known toxins, and can warn scientists when there is insufficient data. Unlike other solutions, this project does not attempt to produce a positive result when there is no confidence in it.
The team has already tested the neural network on more than 126,000 chemical compounds — a volume that would have taken decades using traditional methods. This development can speed up the safety assessment of substances, reduce the number of animal tests, and cut down on routine laboratory work.