We used discrete combinatoric methods and non numerical algorithms [9], based on weighted prefix trees, to examine the data mining of DNA microarray data, in order to capture biological or medical informations and extract new knowledge from these data. We describe hierarchical cluster analysis of DNA microarray data using structure of weighted trees in two manners : classifying the degree of overlap between different microarrays and classifying the degree of expression levels between different genes. These are most efficiently done by finding the characteristic genes and microarrays with the maximum degree of overlap and determining the group of candidate genes suggestive of a pathology.