The identification of meaningful groups of proteins has always been a principal goal in structural and functional genomics. A successful protein clustering can lead to significant insight, both in the evolutionary history of the respective molecules and in the identification of potential functions and interactions of novel sequences. In this work we propose a novel metric for distance evaluation, when applied to protein homology data. The metric is based on a matrix manipulation approach, defining the homology matrix as a form of block diagonal matrix. A first exploratory implementation of the overall process is shown to produce interesting results when using a well explored reference set of genomes. Near future steps include a thorough theoretical validation and comparison against similar approaches.