戻る
J04

Classification of G protein-coupled receptors using self-organizing map

○Takashi Nakayama（Department of Information Sciences, Faculty of Science, Kanagawa University），Akihito Mori（Graduate School of Science, Kanagawa University），Joji Otaki（Department of Biological Sciences, Faculty of Science, Kanagawa University），Haruhiko Yamamoto（Department of Biological Sciences, Faculty of Science, Kanagawa University）

We developed a system for classifying G-protein coupled receptors (GPCRs) to families, and obtained results with high precision for both aligned and non-aligned amino acid sequences. The system developed in this study employs self-organizing map (SOM), which is a kind of artificial neural networks and its output neurons learn to respond individually to distinct input patterns so that nearby neurons respond to similar inputs. That is, SOM performs clustering feature for a given data set. As a result, sequences of amino acids are partitioned to functional family areas (such as amine, peptide, olfactory, etc.) on the map. Since the system predicts families of new sequences and orphans with high precision, it is expected that it works as a strong tool for classifying sequences of amino acids. The learning process were carried out for the various combinations and normalization degrees by using 50 50 square grid neuron space, and the basic batch method. The precisions of the classification are 97~99% for the training set, and 96~98% for the test set. The results suggest that the method works very well for non-aligned data, suggesting that SOM is a promising method for non-aligned data.