Lanthanide-Doped Organic Framework Sensor Array Coupled with Machine Learning for Minimally Invasive Glioma Diagnosis via Cerebrospinal Fluid Biopsy

Glioblastoma multiforme (GBM), the most malignant subtype of glioma, poses significant diagnostic challenges due to limitations in current methods, such as invasive histopathological examination and costly, lab-restricted biomarker detection technologies. Herein, we report a lanthanide (Tb 3+ )-doped organic framework-based sensor array for minimally invasive, sensitive, and accurate glioma diagnosis via cerebrospinal fluid (CSF) biopsy. The sensor array integrates three distinct Tb 3+ -doped frameworks, which exhibit unique topological structures, surface charges, and fluorescence responses, enabling differential recognition of glioma-related biomarkers. The sensor array demonstrated robust discriminatory capacity for eight CSF-relevant molecules via a machine learning algorithm. When applied to clinical CSF samples, it achieved satisfactory separation of glioma patient and normal control samples with 95.5% diagnostic accuracy. This sensor array, combined with advanced machine learning, offers great potential for clinical translation in early glioma diagnosis and molecular stratification.