Using fuzzy memberships to core patterns to interpret connectedness in gene expression clusters

Connectivity based clustering can reveal the continuity of gene expression patterns, and thus can discover interrelations between tumor types, as well as regulatory relations between genes that can lead to discovering gene pathways. Pattern cores are a subset of expression patterns that are representatives of the whole set of patterns and can be used to reveal the structure of the data as well as that of the clusters, especially in the presence of huge data sets. This work presents a fuzzy approach that starts by finding the density-based expression pattern cores. Those cores are then clustered into core clusters and fuzzy memberships to those cores are calculated for all patterns in the data set. The whole data set is then clustered into pattern clusters using a connectivity-based algorithm, where a pattern cluster might contain one or more core clusters. The fuzzy memberships to core clusters in each pattern cluster are used to interpret the connectedness of the pattern cluster using the structure of core clusters, as well as to identify how each pattern is related to one or more tumor types.