Simrank finds the similarity between query sequence A and database sequence B in terms of the number of unique 7-mers that they share, divided by the smallest total unique 7-mer count in either A or B. If you would rather classify a batch of many sequences, use the align and classify tools.