In Silico Drug Repurposing using Knowledge Graph Embeddings for Alzheimer's Disease

Abstract

Drug repurposing (DR), also known as drug repositioning, is a method that identifies novel therapeutic uses from existing drugs. This strategy is highly effective, saves time, cost, and has a minimum risk factor when compared with the traditional in vitro and in vivo drug development methodologies. DR is used for difficult to treat, neglected, or incredibly rare diseases. Alzheimer's disease (AD) is categorized as a difficult to treat disease since, no medication is available that can slow down the disease progression without causing severe risk factors. In this context, novel methodologies are vital to find DR candidates for AD. Natural Language Processing (NLP) is garnering popularity due its ability to discover unseen knowledge and complex associations and this study too utilized a NLP approach. The proposed methodology included four main steps. First, the text data that is needed for the study were retrieved, then the information within the unstructured text data were extracted, next, the DR candidates were predicted using a novel hybrid method which included cosine similarity (CS) and link prediction (LP). Finally, the predicted drug candidates were fed into a trained machine learning (ML) model with a specificity of 0.894% and a f1 score of 0.819% and further validated. The analysis showed that some of the candidates which were not approved nor had ongoing clinical trials have enormous potential in treating AD.