Fig. 5

Overall framework of MVGAEW. A Calculate and integrate the similarities for microbes and diseases. GIP-D represents the Gaussian interaction profile kernel similarity for disease. DSS1 denotes disease semantic similarity while DSS2 denotes disease symptom similarity. GIP-M is similar to GIP-D, DFS1, and DF2 are functional similarities based on disease and drug, respectively. B Adopt an improved VGAE for latent representation with auxiliary tasks. C Utilize XGBoost for potential disease-related microbe prediction by inputting the concatenation of latent representation of each microbe and disease