AI Decodes Cancer Hallmarks: OncoMark Predicts Tumor Behavior from Transcriptomics
Research Summary: We developed OncoMark, an AI framework that predicts the activity of all ten cancer hallmarks from routine tumor transcriptomes, enabling biologically interpretable and clinically actionable cancer profiling.
Researcher Spotlight
Shreyansh Priyadarshi is a Doctoral Researcher (incoming) in the School of Biological Sciences at the University of Southampton. He holds a bachelor’s degree with a major in Biology and a minor in Computer Science, followed by a Postgraduate Diploma in Research from Ashoka University, India. His research lies at the intersection of gene regulatory networks (GRNs), multi-omics data integration, and cellular reprogramming. He is particularly interested in developing computational and systems biology approaches to model regulatory dynamics underlying cell fate decisions.
Personal Website – Shreyansh Priyadarshi
Linkedin – https://www.linkedin.com/in/shreyansh-priyadarshi
Twitter – https://twitter.com/iamspriyadarshi
Lab: Dr. Shubhasis Haldar, S.N. Bose National Centre for Basic Sciences (SNBNCBS)
Lab social media: Home | Structural Mechanobiology Lab
Lab: Dr. Debayan Gupta, Ashoka University
Lab social media: Mphasis AI & Applied Tech Lab at Ashoka – Ashoka University
What was the core problem you aimed to solve with this research?
Cancer is traditionally classified using histology, staging, and individual biomarkers, but these approaches often miss the underlying biological programs that actually drive tumor behavior. The core problem we wanted to address was the lack of a scalable method to directly quantify the complete spectrum of cancer hallmarks from routine biopsy-derived transcriptomic data. Without this, two patients with clinically similar tumors may receive very different outcomes because their tumors are biologically very different.
How did you go about solving this problem?
We approached this challenge by combining large-scale single-cell transcriptomics with neural multi-task learning. We assembled data from nearly 3.1 million cancer cells across 14 tumor types, constructed biologically annotated pseudo-biopsies representing hallmark states, and trained OncoMark to simultaneously predict all ten cancer hallmarks from bulk gene expression profiles. We then rigorously validated the framework across independent patient cohorts and large external datasets to ensure robustness and generalizability.
“OncoMark moves cancer diagnostics beyond appearance, toward understanding the molecular programs that truly define tumor behavior.” – Dr. Shubhasis Haldar
How would you explain your research outcomes (Key findings) to the non-scientific community?
Every cancer has its own “molecular personality”— some tumors grow aggressively, some hide from the immune system, and others become resistant to treatment. OncoMark acts like an AI interpreter that reads a tumor’s genetic activity and identifies which of these hidden behaviors are active. In our validation studies, the system achieved high accuracy and could distinguish cancer from normal tissue while revealing biological changes linked to disease progression.
What are the potential implications of your findings for the field and society?
For researchers, OncoMark provides a new computational framework to study cancer as an integrated biological system rather than isolated pathways. For clinicians, it could support more precise treatment decisions by identifying which hallmark-driven processes dominate a patient’s tumor. In the long term, this could help move oncology from descriptive diagnosis toward truly mechanism-guided precision medicine.
What was the exciting moment during your research?
The most exciting moment was seeing hallmark activity patterns consistently emerge in completely independent patient datasets and realizing that the model was not just making accurate predictions but capturing real biological signals that correlated with clinical tumor progression.
Paper reference: OncoMark: a high-throughput neural multi-task learning framework for comprehensive cancer hallmark quantification | Communications Biology
New Book Launched – Molecules, Mentors & Mindsets: Building Indian Biopharma
Buy your copy today: https://biopatrika.com/science-society/book-molecules-mentors-mindsets-building-indian-biopharma-biocon/
Book Launch: Molecules, Mentors & Mindsets: Building Indian Biopharma | Biocon Focus
Online Interactive Course on Immunology | Starts form June 2026 | Register Today
Online Interactive Course on Immunology | Starts form June 2026 | Register Today
