Two New Techniques To Study The Molecular Mechanism And Clinic Of Epilepsy

ObjectiveChemical genetic technology is used to directionally regulate the most obviously affected nerve nuclei in status epilepticus,so as to seek the origin,transmitters,channels and nuclei of status epilepticus.The endogenous status epilepticus model is established to deepen human understanding of the mechanism of status epilepticus,thus opening up a new field for the study of human status epilepticus.MethodsThe piocapine kindled status epilepticus model was established.The animals were killed 0.5,1,1.5 and 2 hours after the behavioral manifestations.C-fos fluorescence staining was performed,and then the chemical genetic technology was used to regulate the activities of glutamatergic and GABAergic neurons in the most affected nerve nuclei or brain regions in the status epilepticus.As a result,status epilepticus could not be induced.Then the combination of these nuclei and regions was studied according to the principles of proximity(adjacency),similarity(function similarity)and intensity(obvious response after status epilepticus).ResultsIt was found that simultaneous activation of glutamatergic neurons in the hippocampal CA1 area(CA1 area)and ventral anterior thalamic nucleus(VA area)or basolateral amygdaloid nucleus(BLA area)and VA area could induce status epilepticus in animals.This behavior lasted for tens of minutes to hours;Diazepam,midazolam and ketamine can terminate the seizures of animals.The field potential of EEG monitoring is similar to that of human status epilepticus,suggesting that it is an endogenous status epilepticus.The regulation of CA1,BLA and VA regions can not ignite status epilepticus,indicating that the simultaneous activation of the two regions may be a necessary condition for the formation of status epilepticus,in which glutamatergic neurons play a pivotal role.The glutamatergic neurons in the above regions can still be ignited with pilocarpine after necrosis,indicating that its pathogenesis is different from the traditional status epilepticus model.ConclusionCA1 area,BLA area and VA area are the important parts causing status epilepticus.The CO activation of the two sites is a necessary condition for the occurrence of status epilepticus.The activation of glutamatergic neurons that do not meet the physiological needs is the cause of the occurrence of status epilepticus.This endogenous chemical genetic status epilepticus model is closer to human seizures than the traditional model.The success rate of status epilepticus ignited by chemical genetic method is higher,the duration of seizures is longer,and the mortality is lower.It is more suitable for the study of status epilepticus to meet the needs of different experimental purposes.ObjectiveAAbout 30% of epilepsy patients will develop drug-resistant epilepsy(DRE)within a few years after taking the drug.How to identify these patients in the early stage of treatment will help to improve their prognosis,and explore the use of machine learning(ML)in artificial intelligence(AI)to predict the efficacy of antiepileptic drugs after 10 years of treatment.MethodsEight machine learning methods have been developed to predict which patients will suffer anti-seizure medications(ASMs)treatment failure.By training the data of 16 feature dimensions in the early stage of treatment,the accuracy of these methods in predicting treatment failure within next 10 years after medication was evaluated.Randomized grid search method was used to adjust the parameters,and SHapley Additive ex Planations(SHAP)is used as a model interpreter.ResultsFrom the records of 8558 patients collected by the epilepsy center ofthe First Affiliated Hospital of Chongqing Medical University from 2002 to2011,1097 patients who received valproic acid,topiramate or lamotrigine monotherapy and who did not meet the drug resistance criteria were selected as the inclusion criteria.Patients receiving valproic acid,topiramate,or lamotrigine monotherapy and those who did not conform to drug-resistant epilepsy were screened as inclusion criteriaThe Catboost classifier model shows the highest accuracy(74.9%),AUC(78.2%),and F1(58.8%).After parameter adjustment,the accuracy,AUC,and F1 are 76.0%,81.6%,65.7%.Seizure frequency in the first year after medication,maximum daily tablets in one year after medication,and valproic acid were the three most important features contributing to predictive model.ConclusionThe Catboost model can accurately predict risk of ASMs failure in the next ten years by using patients’ response to only one ASM in the first year.This study provide valuable information to improve early identification of DRE patients.