A team of researchers has developed an artificial intelligence framework that shifts how cancer can be interpreted—not by tumour size or spread alone, but by the molecular forces that drive its behaviour. The system, called OncoMark, offers a deeper view into the biological programmes known as the hallmarks of cancer, which determine how tumours grow, evade immunity, resist drugs, and metastasise.
Developed by scientists at the S N Bose National Centre for Basic Sciences in collaboration with Ashoka University, India, the framework analyses 3.1 million single cells from 14 cancer types to create synthetic “pseudo-biopsies.” These models capture hallmark-driven disease states, allowing the AI to learn how malignant programmes interact and intensify as cancer progresses.
In internal tests, OncoMark reached more than 99% accuracy, and remained above 96% across five independent cohorts. It was validated on 20,000 patient samples from eight major datasets—one of the broadest tests of its kind. For the first time, researchers could clearly visualise hallmark activity rising with advancing tumour stage.
Published in Communications Biology, the system can reveal which cancer hallmarks dominate in an individual patient, helping clinicians identify aggressive disease earlier and select therapies that specifically target the underlying biological mechanisms driving tumour growth.
Author
