Integrating machine learning and retrospective cohort to explore the effects of AEDs on the prognosis of LGG and identify related molecular targets.

Abstract

Antiepileptic drugs (AEDs) were frequently employed in glioma patients, especially those with low-grade glioma (LGG), in which epilepsy manifested in roughly 70-90% of cases. It has been reported that some AEDs can improve the survival of glioma patients. However, the molecular mechanisms of AEDs through which affect LGG prognosis remained unclear. Therefore, this study integrated 105 targets of 10 AEDs, by using machine learning, molecular docking, a retrospective clinical cohort, and in vitro experiments to clarify the biological mechanisms through which AEDs affect LGG prognosis. Our study established a reliable 13-gene AEDs-related LGG prognostic model. Carbamazepine, Oxcarbazepine and Lacosamide were supposed to improve the OS by inhibiting the hazard factor SCN9A. Phenobarbital was supposed to restrict the OS by inhibiting the identified protectors GRIN2C and GRIN3A. Molecular docking visualized the strong affinities between the above drugs and targets. Retrospective cohort further verified our speculation about the effect of the above drugs on the prognosis of LGG. In vitro experiments demonstrated that inhibiting SCN9A, a common targets for most AEDs, could suppress LGG cells malignant behaviors; while the suppression of GRIN2C and GRIN3A enhanced the malignant behaviors of LGG cells. This study provided guidance for individualized AEDs selection for LGG patients, and provide new insights into the potential biological functions and molecular mechanisms of AEDs affecting LGG.