Gene Ontology Semi-supervised Possibilistic Clustering of Gene Expression Data

Clustering is one of the most important data analysis methods with applications of significant importance in many scientific fields. In computational biology, clustering of gene expression data from microarrays assists biologists to investigate uncharacterized genes by identifying biologically relevant groups of genes. Semi-supervised clustering algorithms have proven to bring substantial improvements in the results of standard clustering methods especially on datasets of increased complexity. In this paper we propose a semi-supervised possibilistic clustering algorithm (SSPCA) utilizing supervision via pair-wise constraints indicating whether a pair of patterns should belong to the same cluster or not. Furthermore we show how external sources of biological information like gene ontology data can provide constraints to guide the clustering process of SSPCA. Our results show that the proposed algorithm outperformed other well established standard and semi-supervised methodologies.