This project evaluates the feasibility of artificial intelligence methods to model longitudinal patient electronic health records (EHRs) to favor the prediction of neurological disorders and comorbidities, the recommendation of the appropriate treatments for sick patients and the discovery of potentially alternative therapeutics to prevent cognitive decline. Therefore we propose and evaluate a largely novel approach based on deep learning for comprehensive modeling of neurological disorders, in particular dementia and Alzheimer’s disease, from patient EHRs to leverage clinical decision-making, disease phenotyping and drug discovery.