Machine Learning techniques to identify potential drug targets for Anti-epileptic drugs

Epilepsy is a neurological disorder affecting millions worldwide. Though many Anti-epileptic drugs (AEDs) are available for treatment, almost 30% of patients with Epilepsy (PWE) are resistant to these AEDs. In this paper, we applied Machine learning (ML) techniques for predicting potential drug targets for development of new AEDs. For this particular problem statement, some of the widely used classification algorithms have been applied. Features related to physico-chemical, structural properties, and post-translational modifications are considered for training the ML models. These models have been trained using three different datasets i.e. epilepsy-associated proteins known from literature, all the reviewed human proteins, and known AED targets. Applying different approaches, an accuracy of more than 80% is achieved, and even the F1 score is found to be significant. We have identified few novel druggable proteins that could act as potential targets for patients with refractory seizures.