Abstract
We developed a computational method to detect identities in tRNA genes. The method uses the multidimensional scaling method to classify the sequences of tRNA genes into multiple groups of similar sequences, and also to extract characteristic bases that are conserved within a group but differ from other groups. This procedure was applied recursively to classify the sequences into hierarchical groups so that characteristic sites can be detected more precisely. We were able to detect many characteristic sites in T and D domains of tRNAs as well as the characteristic sites that have been detected experimentally. This suggests that the preservation of L-shape structure in tRNAs is important to the tRNA-ARS recognition.
