In order to find the function of genes from gene-expression profiles, the hierarchical clustering has been used generally. But this method has problems, for example a dendrogram tends to change by data dependence, therefore it is easy to be influenced of the error of an experimental noise.

To cope with the problems, we propose another type of clustering. We formulate the problem of the clustering as a graph-covering problem by connected subgraphs where vertices and edges of the graph denote genes and similarities between genes, respectively. The method is based on the p-quasi complete linkage algorithm for describing clusters. We present the outline of an algorithm for clustering a set of genes into subsets corresponding to p-quasi complete linkage graphs.