Fig. 3
From: Predict multi-type drug–drug interactions in cold start scenario
Computing framework of CSMDDI model. Embedding learning will learn the embeddings of the all existing drugs and DDI types. Mapping function learning will learn a mapping function between DDI network topology and drug’s attribute. Prediction will use the learned embedding vectors of a drug pair (one is existing drug and another is new drug, or two of them are new drugs) as the input, and then choose a predictor to output their prediction score. A higher score indicates a higher probability of this interaction occurrence