Fig. 5.

The dimensional reduction visualizes the latent vectors learned by mpDisNet. The latent vectors learned by mpDisNet by combining M1 (disease–miRNA–gene–gene–miRNA–disease) and M3 (disease–gene–gene–disease) meta-paths on an integrated network of disease–gene and disease–miRNA–gene (Fig. 1). We only illustrated the diseases with the well-defined pathobiological category with at least seven types of diseases. The diseases are classified according to the clinically annotated pathobiological classification data (color key) from a previous study.¹⁶