St. Petersburg chemists taught AI to search for nanoparticles that destroy only cancer cells
ITMO has developed an AI-based platform to search for nanoparticles with selective toxicity to cancer cells. The study, published in the journal Small, Hitech, was reported by the press service of ITMO University.Researchers from the Center for Artificial Intelligence in Chemistry of the SCAMT science cluster at ITMO University have trained an AI to predict the toxicity of nanoparticles for cancer and healthy cells. The algorithm selects particles that target only diseased cells, reducing side effects of therapy.
The researchers use a trained model for prediction that estimates the toxicity of nanoparticles based on size, charge, composition and surface characteristics. The AI uses an algorithm that mimics evolutionary selection to pick nanoparticles that are toxic only to cancer cells. The system generates variants of new nanoparticles and also evaluates the effectiveness of existing ones to target specific tumors.
Machine learning is already being used to assess the toxicity of nanoparticles, the researchers note. The key innovation of the new solution is the genetic algorithm. Simulating evolutionary selection speeds up the selection of the right toxin. In addition, the system evaluates the degree of toxicity of specific nanoparticles for specific cells.
Inorganic nanoparticles are used to deliver drugs to cancer cells. They reduce the negative effects of the drug on healthy cells, increase drug solubility and even destroy the tumor on their own. But some particles are toxic not only to diseased cells, but also to healthy cells.
Using AI will speed up the selection of particles for targeted therapies, reducing the time and cost of trials. The researchers are working on adapting the algorithm to find particles toxic to certain bacteria and fungi that cause infections.