Manual specification of network communities

[wrmartin]

I’m working with a ~300 residue protein, and I’d like to use communities for network analysis. However, I have 24 forms I’m comparing, and the communities understandably do not align. What I’d like to do is specify ~10 communities; as an example, every 30 residues is one community (not what I would actually use). Basically, I want to use the network analysis to determine the strength of interaction of specific regions (mainly defined by secondary structure).

[Xinqiu Yao]

Hi,

The traditional community analysis using maximal modularity has a known problem of community comparison. There are multiple ways to alleviate the issue. One is exploring the modularity vs # of communities plot and choosing the simplest partition close to the maximal. This hopefully merges small communities that do not contribute to modularity. You can also manually adjust the definition of communities (such as choosing the boundary of secondary structures) to best match your systems. An example application can be found in the paper: https://doi.org/10.1074/jbc.M115.702605 Example code to do the analysis is available upon request to [e-mail]

‌

The other way you may think of is using our recent dCNA method, which generates a consistent community partition across multiple systems, allows flexible adjustment of the number of communities, and produce more meaningful results for interpreting community-community interactions. See the paper: https://doi.org/10.1021/acs.jcim.8b00250. Code is available at https://github.com/The-Hamelberg-Group/dcna

‌

Hope it may help.

‌