Figure 4
Communities’ analysis methodology.
(A) Schematic representation of spontaneous activity traces for 5 clusters. All those clusters that fire concurrently in a short time window define a sequence (pink bars). Clusters that fire independently are discarded. Sequences shape the X matrix, where each column corresponds to a sequence and each row represents the activity history of a cluster. Two or more clusters that share a similar history are more likely bound and would constitute a community. The degree of similarity between all pairs of clusters (i, j) is established through the Jaccard’s similarity measure JS(i, j), from which the Jaccard’s matrix distance JD = 1 − JS is determined. Clusters #1 and #2 are identical in history and provide JS(1, 2) = 1, but both are also similar to #5 (yellow bands), with JS(1, 5) = JS(2, 5) = 2/3. (B) JD is a symmetric matrix that reflects the relative closeness of all pairs of clusters, which can be sketched as spatial groups or in the form of a dendrogram. Clusters #3 and #4 have identical histories and form a unique community. Clusters #1 and #2 also shape a community, but they are sufficiently close to cluster #5 to constitute together a higher, more representative group. The number of communities is formally set by selecting a threshold in the dendrogram. Any threshold along the red arrow would maintain two communities. Once a threshold is set, the similarity matrix is ordered to visually highlight the communities in the network.
