Protein Interaction Networks (PINs) are subject to biases both in their construction (as better studied proteins are more likely to be highly connected) and in the detection of functional modules (due to biases in the ways these proteins are studied). These biases can make it more difficult to find useful modules that are functionally relevant and not just disproportionately understood due to trends in research. One way to address this bias is to repeatedly sample the local network background of a candidate community using a random walk. CommWalker is a method that tries to obtain an impression of the level of local functional homogeneity of a community (Leucken et al, 2017). By accounting for these background levels of functional similarity within a network, communities that are truly, uniquely functionally cohesive can be brought to the fore. These communities may be just as interesting as better studied parts of the network but may not yet have been thoroughly investigated, creating the potential for novel biomedical discoveries!
<br><br>
Luecken MD, Page, MJT, Crosby AJ, Mason S, Reinert G, Deane CM, CommWalker: Correctly Evaluating Modules in Molecular Networks in Light of Annotation Bias, Bioinformatics, 2017