On the validation of gene expression clusters

Classical validity indices are limited to clusters of specific geometrical shapes, and do not allow for discovering the natural cluster structure in data. Moving from the traditional coherent gene expression clustering to exploring the connectivity of gene expression patterns demands the use of more efficient validity indices. In this work, the application of a novel validity measure to gene expression clustering is investigated. The measure was previously introduced by the author to evaluate clusters of arbitrary shapes in low and high dimensional data. However, experimenting with its applicability to gene expression data is the main focus of the presented work. Leukaemia and breast cancer sets are used to evaluate the applicability of the proposed validity to discover the natural connectedness of expression patterns.