Validation of biomarkers predicting clinical outcomes of umbilical cord blood transplantation

Umbilical cord blood transplants often fail because the donated stem cells are too slow to rebuild a patient's blood and immune system. Researchers will analyse hundreds of cord blood units, using machine learning to identify molecular markers that predict which units will engraft quickly and reliably in both lab animals and real patients. Blood cancers and inherited blood disorders such as sickle cell disease can be cured by a stem cell transplant, but many patients never find a matching adult donor. Cord blood banks hold thousands of ready-to-use units, yet roughly one in three transplants results in delayed or failed engraftment, leaving patients vulnerable to infection and bleeding for months. No reliable test currently exists to pick the best unit for a given patient. If the team can validate a set of predictive biomarkers, hospitals could screen cord blood units before transplant and select only those with the highest chance of success. That would reduce the number of failed transplants, shorten hospital stays, and make cord blood a more dependable option for patients who have no other donor. The machine-learning tool developed here could eventually be packaged as a clinical decision-support system for transplant centres worldwide.