Scientists have adapted AI to treat rare cancers

Scientists have adapted artificial intelligence (AI) to treat rare cancers. Dmitry Chebanov, the author of the idea and implementation of the project, Doctor of Sciences, systems biologist and researcher at the Memorial Sloan Kettering Cancer Center (MSKCC), told Izvestia on April 29. The results of the research into a system for treating cancer by creating digital patients were presented at the annual conference of the American Association for Cancer Research (AACR Annual Meeting).

The scientific guidance was provided by Professor Quaid Morris, Head of the Computational Biology and Medicine Program at Cornell University and a world expert in the field of genomics and tumor evolution.

As Chebanov explained, a new artificial intelligence-based system has been developed that not only analyzes data on cancer patients, but creates their virtual digital counterparts to fill in the gaps in knowledge about rare tumors. According to him, the technology opens up new opportunities for drug development and could change the practice of personalized medicine in the coming years.

"AI is usually used in medicine to find the necessary information in electronic patient records. But a fundamentally new approach has been applied here: the model creates new molecular profiles by itself, rather than just reading existing ones. This is the first time such technology has been used to generate synthetic cancer data," said the author of the study.

In concert with sounds: AI will customize children's hearing implants in minutes

How a neural network will help speed up the adaptation of young patients with hearing loss

Data on molecular disorders in tumor cells became the basis for the choice of therapy. However, as the systems biologist noted, such information is not enough for rare cancers: there are few patients and samples. It may take years to accumulate statistics, but AI can generate the missing data now.

The researchers proposed using the principles of large language models by processing the molecular profiles of tumors as "sentences" and creating hundreds or thousands of digital counterparts of real patients based on them. To test the quality of the technology, the scientists trained a new diagnostic classifier on virtual patients and showed that its accuracy in making diagnoses in real people exceeds that of a similar model trained on real data. The development also makes it possible to quickly supplement incomplete molecular profiles, which is important for a quick start of therapy.

Earlier, on February 14, Andrey Varivoda, Chairman of the Board of Directors of the DEACON Group of Companies, told Izvestia that mass implementation of AI systems in regional medical institutions is expected by 2026. He also clarified that more than 30 AI-based services for analyzing medical images are known in Russia. The systems help doctors to detect oncological diseases, pathologies of the lungs and the cardiovascular system at an early stage.