Researchers map brain cancer gene regulation with million-cell AI analysis

Researchers publishing in Nature developed scDORI, a deep-learning tool that maps how genes regulate cell identity, and applied it to profiles of more than one million cells from primary glioblastoma tumours. The method captures continuous differences between cells without requiring predefined cell categories and models both gene activation and repression. In experiments using patient-derived cancer cells, 76% of tested regulators produced identity shifts consistent with the tool's predictions. The researchers also identified MYT1L as a regulator that constrains changes in cancer cell identity; increasing its activity reduced tumour growth after transplantation in mice. Those therapeutic findings remain preclinical.