UC San Diego researchers described CANDiT, a machine‑learning systems biology framework that identifies targets to reprogram cancer stem cells toward differentiation and cell death. Applied to colorectal cancer models, the tool prioritized networks restoring CDX2 expression and suggested interventions that selectively affect cancer cells while sparing normal tissue; authors published results in Cell Reports Medicine. The approach revives the concept of differentiation therapy for carcinomas and offers a data‑driven path to target elusive stem‑like tumor cells across cancer types. Investigators framed CANDiT as tumor‑specific: it starts from a key gene and maps networks to actionable targets, potentially shortening the route from discovery to trial.