Why should we remove the self-connected proteins when do gene prioritization work

Question

[baiguihuajl]

Why should remove self-linked proteins in protein-protein interaction network in gene prioritization work, such as in this work: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5374555/

Answer 1

[Jean-Karim Heriche]

It depends on the context. Self interactions in a graph affect node degrees. This for example slows down random walks on the graph. Often the edge weights represent some form of similarity and self similarities tends to be higher than similarities between different nodes which may create a bias in data retrieval. Also, it doesn't always make biological sense to assume that a protein interacts with itself.