UCSD Scientists Use AI to Destroy Cancer Stem Cells
Researchers at the University of California, San Diego (UCSD) have developed an AI-driven framework called CANDiT (Cancer Associated Nodes for Differentiation Targeting) to selectively eliminate cancer stem cells (CSCs)— the resilient cells responsible for tumor recurrence, metastasis, and treatment resistance.
CANDiT uses machine learning and systems biology to identify therapeutic targets that can reprogram CSCs, prompting them to behave like normal cells and ultimately self-destruct. Unlike conventional therapies, it targets only cancer cells without harming healthy tissue.
In their study, the team applied CANDiT to colon cancer, focusing on restoring the CDX2 gene, whose loss is linked to poor outcomes. By screening over 4,600 human tumors, the system identified PRKAB1 as a key target. Using the experimental drug PF-06409577, a PRKAB1 activator already shown to be safe in humans, researchers successfully re-expressed CDX2 in colon cancer stem cells. Unexpectedly, once reprogrammed, these cells chose to die naturally, losing their ability to survive without their cancerous state.
The approach was validated using patient-derived organoids (PDOs) at UCSD’s HUMANOID Center, allowing researchers to simulate human clinical responses “in a dish.” Using computational trials on data from over 2,100 patients, they found that restoring CDX2 could cut recurrence and mortality by up to 50%.
Beyond colon cancer, CANDiT offers a scalable, tumor-agnostic platform for identifying precision differentiation therapies across multiple solid cancers. The team is now optimizing the compound with chemist Jerry Yang, PhD, and expanding trials to other tumor types under Michael Bouvet, MD.
According to senior author Pradipta Ghosh, MD, this marks a major step toward targeting and eradicating cancer stem cells, potentially reshaping how solid tumors are treated.
Credit: GEN News