Indian Researchers Create OncoMark AI System for Personalised Cancer Treatment

Indian Scientists Introduce OncoMark — AI System for Personalised Cancer Care

Indian scientists have announced a breakthrough in cancer research with the development of OncoMark , an advanced AI-based analytical framework designed to personalise cancer diagnosis and treatment. The innovation has emerged from collaborative efforts between the S N Bose National Centre for Basic Sciences and Ashoka University , opening new possibilities for molecular-level assessment of tumours.

Unlike conventional staging methods that classify cancer primarily by size, spread and physical characteristics, OncoMark examines biological hallmarks — the core molecular mechanisms that turn cells cancerous , promote metastasis, enable immune evasion and drive resistance to therapy. This shift from stage-based evaluation to process-based characterisation allows clinicians to understand how and why tumours behave differently , even among patients diagnosed at the same stage.

Massive Dataset & Pseudo-Biopsy Modelling

Scientists trained the framework using a large database of 3.1 million tumour cells representing 14 cancer types . They created synthetic pseudo-biopsies to map hallmark-linked tumour states, enabling the algorithm to track progression patterns such as genomic instability and immune escape. The model achieved >99% internal prediction accuracy , and over 96% accuracy across multiple independent datasets , including 20,000 real patient samples from eight global studies .

Transforming Cancer Treatment & Early Detection

By identifying which hallmarks are active in a tumour, oncologists can select targeted therapies rather than relying on uniform treatment protocols. This could improve prognosis, detect aggressive cancers earlier, and provide personalised care strategies — a major advancement in precision medicine.

Exam-Point Quick Facts

• AI tool name: OncoMark , developed jointly by SN Bose Centre & Ashoka University

• Data used: 3.1 million single cells across 14 cancers

• Accuracy: >99% internal, 96% independent datasets

• Published in Communications Biology (Nature Group)