Scientists Develop ‘OncoMark’, a New AI Framework for Cancer Behaviour Prediction

Researchers have developed a new artificial intelligence framework, OncoMark, designed to analyse the behaviour of cancer cells with high precision. The model was trained on genetic information from 3.1 million cells representing 14 types of cancer. This extensive training enables OncoMark to assess core biological mechanisms, such as immune evasion, metastasis, and other hallmark processes that contribute to tumour progression and treatment resistance.

When evaluated across five independent datasets, OncoMark demonstrated consistent performance, achieving an accuracy rate of 96 per cent. According to the research team, the next step is to integrate the framework into clinical workflows, making it available to clinicians for large-scale use in diagnosis and treatment planning.

The team also plans to extend the model to blood cancers, which exhibit growth patterns distinct from solid tumours, as well as to other rare cancer types that currently lack comprehensive predictive tools.

The OncoMark framework was developed by scientists from Ashoka University, Sonipat (Haryana), and the S. N. Bose National Centre for Basic Sciences, Kolkata.

The work is detailed in a research paper published in October 2025 in Communications Biology, which outlines how OncoMark uses machine learning to analyse hallmark cancer processes using genetic data.

Source: indiatimes.com