Williamson CM, Franks B, Curley JP, Mouse Social Network Dynamics and Community Structure are Associated with Plasticity-Related Brain Gene Expression, Frontiers in Behavioral Neuroscience.



Link to Paper Online

We have previously shown that twelve male mice living in a visible burrow systems form social hierarchies. The aim of this study was to characterize the social network dynamics of a large group of male mice and to evaluate how individual network position relates to hippocampal gene expression. We housed 30 male mice in four large interconnected vivaria. Vivaria 1&2 were connected between two nestboxes with tubeA. Vivaria 3&4 were connected between two nestboxes with tubeC. Vivaria 2&3 were connected via the long tubeB. We observed animals for two hours per day for 19 days recording all occurrences of dominant/subordinate behavior that occurred between any pair of animals. vivaria

We found that male mice formed a highly significant linear social hierarchy and we could rank animals successfully from 1-30 using the I&SI method. We also report in the paper on the close relationships between I&SI rank and individual network measures such as hub score, out- and in-closeness, and out- and in-degree.

The matrix below on the left below shows the raw data. Numbers in cells represents the number of times that the animal in each row won a fight against animals in columns. In the matrix below on the right we binarize the left matrix to a 1/0 with a 1 representing a winner in each relationship (the animal in the row). The I&SI matrix attempts to linearly order winners and losers according to the minimum number of inconsistencies in rank order. Animals are listed in their I&SI rank order. Hover over the matrix to compare wins/losses between two animals.

Win/Loss Sociomatrix

Binarized Sociomatrix


A key finding of the study was that mice did not use space equally. Dominant individuals were more likely to localize their behavior over time to particular vivaria. Even subordinate mice became more localized over time. This led to mice associating more frequently with specific other mice. We formally determined that the mice formed two major network communities with each being hierarchically organized and containing an alpha male. Most mice spent the majority of time in one of these network communities, but some mice would rapidly switch between. For this analysis we use the Girvan & Newman community detection algorithm, boostrapping our data 1000 times to generate confidence in the network community associations. Details are available in the paper. The figure below shows the strength of relationship between each pair of individuals based on the community detection method. Numbers refer to the I&SI rank. network



A three dimensional interactive visualization of this network is available here

Another key finding from this paper was that the behavior of animals on standard laboratory behavioral tests of exploration and social behavior prior to group formation was not at all predictive of their final social hierarchy/network position. Interestingly, it did predict some early behavior in the group but not later behavior. Secondly, we observe that social network position / social rank is associated with hippocampal DNMT1 mRNA expression. We found that less powerful/dominant individuals have higher DNMT1 expression, perhaps suggesting a decrease in overall gene expression.

We are interested in determining what the long-term effects of living in such complex social networks have on brain physiology, and how the brain rapidly adapts to changes in the social environment to promote socially competent behavior.