Clustering biological data by unraveling hidden transitive substructures