Table 1 Summary of four particular clustering approaches.
From: Microbial community pattern detection in human body habitats via ensemble clustering framework
Clustering Approaches | Characteristics | Limitations on microbial pattern |
|---|---|---|
Density-based clustering | Clusters are defined as connected dense regions in the network | True microbial community are not limited to densely connected structures; sparsely microbial structure still exists |
Graph partition-based clustering | Clusters are generated via graph partitioning techniques | Partition based algorithms do not allow the overlaps between clusters. Therefore, they are unable to discover shared microbe among clusters, such as some species that could adapt in multi-environmental conditions like microbial mats and biofilms |
Hierarchical clustering | Clusters are built based on an agglomerative clustering model that shows relations between the members and groups | Hierarchical structure is determined by local optimization criterion as such there is no global objective function, which might lead to small clusters with only part of similar samples |
Distribution-based clustering | Clusters are modelled using statistical distributions | Statistical models of microbial communities are still unknown and need to be further explored |