Figure 5.

The flowchart of design and discovery of a multi-molecule drug for therapeutic treatment of psoriasis. Drug-target interaction data were obtained from drug-target interaction databases. Then, the drug and target feature vectors were pre-processed, including downsampling, standardization, and PCA, respectively. After data preprocessing, drug target feature vectors were divided into training data and testing data for training the DNN-based DTI model. The well-trained DNN-based DTI model was used to predict candidate drugs for these biomarkers (drug targets). The potential molecule drugs were selected from predicted candidate drugs according to the drug design specifications and combined as multi-molecule drugs for the therapeutic treatment of psoriasis.