Network Diffusion




Network Diffusion (nDiffusion) application can robustly and accurately prioritize and validate functionally related genes, facilitating functional interpretations of large omics data and biological knowledge discovery. nDiffusion application website can answer the following question:

Are two groups of genes functionally related to each other?

To answer this, nDiffusion works under the assumption that genes that are highly connected to each other in biological networks are more likely to participate in the same biological processes and share similar functions. nDiffusion utilizes Graph-based Information Diffusion (GID) method {PMID: 21179190 , 25126794} in order to evaluate how well two groups of genes are connected to each other in a network. GID simulates the flow of liquid or information, starting from seed nodes with certain information or known functional annotations, and spreading the information throughout the network to other nodes. Nodes that are closer to the starting nodes, meaning that they are few edges away and the edges have higher confidence weights, will receive more information signals and thus, more likely to share similar functions.

GID has been shown to prioritize genes associated with the same pathways, ontologies, and clinical phenotypes faster and more accurately than the conventional shortest path length approaches {PMID: 31797617}. This is because the diffusion method simultaneously considers both edge confidence weights and multiple paths that genes are connected to each other in the networks, while the shortest path length can only do deduce paths with the smallest number of steps or the smallest sum of edge weights.

Please refer to this presentation for more detailed explanations.