Consensus Clustering - Tunable-tightness Partitions

Tunable-tightness Partitions

In this different form of clustering, each data object is allowed to be exclusively assigned to one and only one cluster, to be unassigned from all clusters, or to be simultaneously assigned to multiple clusters, in a completely tunable way. In some applications like gene clustering, this matches the biological reality that many of the genes considered for clustering in a particular gene discovery study might be irrelevant to the case of study in hand and should be ideally not assigned to any of the output clusters, moreover, any single gene can be participating in multiple processes and would be useful to be included in multiple clusters simultaneously. This has been proposed in the recent method of the binarization of consensus partition matrices (Bi-CoPaM) and is being used currently in the field of bioinformatics.