Fig. 1
From: Prediction of adverse drug reactions based on knowledge graph embedding
Overview of the work flow of this study. a Knowledge graph composed of the drug, targets, indications, and side effects extracted from the DrugBank and SIDER databases; b The knowledge graph embedding process, (b-top) Word2Vec training corpus constructed based on the knowledge graph; (b-middle) Continuous bag-of-words (CBOW) implementation process of Word2Vec, where the input layer inputs any two elements in the triple, the other element is used as the output (represented by one-hot vector), and W is the vector matrix of the training elements (entities and relations); (b-bottom) vector matrix of the training elements, W; and c Binaryclassifier, with the vector difference of the drug and side effect pair as the input and the probability that the drug may cause the side effect as the output