Fig. 1

Graphical overview of PWN. PWN generates a weighted and implicitly directed network (lower right) from an unweighted and undirected network (upper left) using two distinct sources. First, PWN computes the Ricci curvature of the edges and derives the first edge weights by applying an exponential function on the computed curvature (upper middle). We consider this curvature an internal feature. After that, the external feature warps the network; prior knowledge is mapped on the network nodes (upper right) and then spread. The spread prior knowledge is then applied to the edges (lower right). Finally, gene scores acquired from the omics data is mapped on the nodes (lower middle) and spread to obtain the final scores (lower left)