Новый ИИ-инструмент предсказывает болезни на десятилетия вперёд Translation: New AI Tool Predicts Diseases Decades Ahead

Scientists have created an AI tool capable of predicting over 1,000 diseases and forecasting changes in health status up to a decade in advance.

Experts from the European Molecular Biology Laboratory (EMBL), the German Cancer Research Center, and the University of Copenhagen employed algorithmic principles similar to those found in large language models.

The AI was trained on data from two independent healthcare systems—using anonymized information from 400,000 individuals participating in the UK Biobank study and 1.9 million patients from Denmark’s national registry.

«Medical events often follow predictable patterns. Our AI model analyzes these patterns and can forecast future health outcomes,» said Thomas Fitzgerald, a researcher at the European Bioinformatics Institute.

The new tool assesses the likelihood of an individual developing specific diseases and when this may occur. The neural network can predict cancer, diabetes, cardiovascular conditions, respiratory illnesses, and numerous other disorders.

The Delphi-2M model examines «medical events» in a patient’s history and lifestyle factors, including weight issues, harmful habits, age, and gender.

Health risks are expressed in percentage terms over time, akin to weather forecasts: «There is a 70% chance of rain this weekend.»

Euan Birney, the acting director of EMBL, mentioned that patients will start benefiting from this tool in the coming years:

«You come in for a consultation, and the doctor is already using these tools, saying, ‘Here are the four main risks in your future, and here are two things you can do to change this.'»

He acknowledged that standard advice like losing weight or quitting smoking remains important, but for certain diseases, there will be more specific recommendations.

Birney highlighted that a key advantage of Delphi-2M over other solutions is its capability to predict all diseases simultaneously over an extended timeframe.

«Delphi-2M assesses the likelihood of more than 1,000 diseases based on an individual’s medical history, and its accuracy is comparable to existing models focused on single diseases,» the project team stated.

Professor Moritz Gerstung, head of the AI department at the German Cancer Research Center, emphasized that Delphi-2M marks the beginning of a new approach to understanding human health and disease progression. He stated that generative models could eventually personalize care and anticipate healthcare needs on a system-wide scale.

Additionally, in September, researchers from Harvard Medical School introduced an AI model capable of identifying precise combinations of genes and drugs to reverse pathological conditions in human cells.