Impact of score and network properties on the prioritization.

<p>Subfigure A) shows the impact of the prioritization score properties. Subfigure B) shows the impact of network properties. Disease genes are depicted in red, light blue and orange. Edges between a disease gene and its neighbors have the same color as the disease gene node. Subfigure A) pictures the paths considered when computing the score for a given candidate (green box) using full topology, shortest paths, direct neighbor overlap (EndNet), and direct neighbor approaches, respectively. Edges in grey are ignored. Full topology is the most comprehensive strategy, as it considers all paths. In shortest paths, the alternative (shortest) path between the candidate and the orange gene (dashed line) is not taken into account. Contribution of additional paths of different lengths is also ignored. Using the direct neighbor overlap, the candidate can still be linked to the disease, but with no contribution from the orange gene. With direct neighbors, the candidate receives a score of 0. Subfigure B) shows the effect of dense <i>vs</i> sparse interactions and good <i>vs</i> poor genome coverage. Heterogeneous associations (left) <i>vs</i> single source interactions (right), the latter with absent genes (blue gene) and lack of connectivity (red gene).