A Structural Data Mining Approach for the Classification of Secondary RNA structure

There exist many methods for classifying genomic data by aligning, comparing, and analyzing primary nucleotide sequences using such algorithms as dynamic programming and kinetic folding, etc.. These methods are, however, not always effective as motifs are more conserved in structures than in sequences. Instead of performing classification based on primary sequences, we therefore propose to perform the task from structure, exploiting the phenomenon in which molecules form from a sequence of nucleotides, beginning with a primary sequence that can fold back onto itself to form a secondary structure and then a tertiary structure. The algorithm we propose is able to perform data mining in structural data and is called the random multi-level attributed (RMLA) graph algorithm for mining and representing secondary genomic structure from such biomolecule as tRNA. The identification of structural similarity is implemented with information measure concept to characterize the resultant class. Experiments are based on known tRNA structural data from database of compilation of tRNA genes. The results show that our approach is able to effectively classify different class of tRNA secondary structure. We also compare our result with other classification algorithms to prove the effectiveness. Our approach shows a better way to classify structural data. In fact, RMLA graph is not suitable only for the classification of genomic data, wherever graphs are used to model data, it is useful for discovering patterns in the databases