Indian Researchers Develop AI Tool to Detect Hidden Cancer Stem Cells

Led by Dr. Shubhasis Haldar, the research team developed ACSCeND, short for AI-based Cancer Stem-like Cell Profiler and Neoplasm Deconvoluter.

Indian researchers have developed an artificial intelligence framework capable of identifying hidden cancer stem-like cells within large tumour datasets, opening a possible route towards more precise assessment of cancer recurrence, treatment resistance and patient outcomes. Developed by scientists at the S. N. Bose National Centre for Basic Sciences (SNBNCBS) in collaboration with Ashoka University, the system could prove particularly useful when advanced single-cell sequencing facilities are not readily available.

Cancer treatments can eliminate vast numbers of tumour cells, yet small populations of stem-like cells may survive and contribute to the cancer returning, spreading to other organs or becoming resistant to therapy. Detecting these cells has remained difficult because they are rare and can change their biological identity as tumours develop.

ACSCeND Identifies Three Cancer Stem-Like Cell States

Led by Dr. Shubhasis Haldar, the research team developed ACSCeND, short for AI-based Cancer Stem-like Cell Profiler and Neoplasm Deconvoluter. Unlike conventional approaches that give a tumour a single "stemness" score, ACSCeND identifies three developmental states described as pluripotent-like, multipotent-like and unipotent-like, giving researchers a more detailed picture of differences within a tumour.

The framework combines information learned from high-resolution single-cell sequencing with deep learning and applies that knowledge to conventional bulk tumour RNA-sequencing data. This approach makes it possible to investigate hidden cell populations across large numbers of patient samples without requiring single-cell experiments for every tumour.

More Than 25,000 Tumour Samples Analysed

Researchers validated ACSCeND against existing computational approaches across independent datasets and sequencing platforms before using it to examine more than 25,000 tumour samples from major international cancer databases, including TCGA and PRECOG.

The analysis found that tumours containing higher levels of pluripotent-like cancer stem cells were linked with poorer patient survival, increased chances of tumour recurrence and weaker responses to modern immunotherapies, indicating that the developmental state of stem-like cells may provide valuable information about how aggressively a tumour could behave.

New Framework Builds on OncoMark AI Platform

ACSCeND builds on the team's earlier AI platform, OncoMark, which was developed to decode biological hallmarks involved in cancer progression across millions of cells and reported predictive accuracy of more than 99%. The latest work takes that approach further by concentrating on the rare stem-like populations believed to play an important role in tumour evolution.

AI Could Support More Precise Cancer Treatment

The framework also identified molecular programmes that may help cancer stem-like cells survive, adapt and escape immune responses. Such information could support future research into new drug targets while helping scientists understand which patients may face a greater risk of relapse or reduced treatment response.

Tools such as ACSCeND could make complex genomic information more accessible for cancer research, including in settings where advanced laboratory facilities are limited, while supporting the broader development of precision cancer medicine.