Drug repurposing prediction for immune-mediated cutaneous diseases using a word-embedding based machine learning approach

Immune-mediated diseases affect >20% of the population and many autoimmune diseases affect the skin. Drug repurposing (aka repositioning) is a cost-effective approach for revealing drugs that can be used to treat diseases for which they are currently not prescribed. We implemented an efficient bioinformatics approach using "word embedding" to summarize drug information from >20 million articles, and applied machine learning to model the drug-disease relationship. We trained our drug repurposing approach separately on 9 cutaneous diseases (including psoriasis, atopic dermatitis and alopecia areata), as well as 8 other immune-mediated diseases, and obtained a mean AUROC of 0.93 in cross-validation.