Protein motif çıkarımında öğrenme-tabanlı yaklaşımlar

Motif extraction is a challenging problem in bioinformatics. The reason for this arises from the importance in identifiying and predicting the structural and functional regions in the biological sequences. Various statistical methods and machine learning-based approaches could be used for motif extraction. Neural fuzzy systems are hybrid structures where fuzzy systems have the learning ability of artificial neural networks. Adaptive neural fuzzy inference systems (ANFIS) are integrated systems in which fuzzy concepts are applied in adaptive neural networks. In this thesis, learning-based approaches are on the motif extraction are investigated. A method has been developed to extract motifs from the Manganese and Iron SuperOxide Dismutase enzyme family. Method has the following stages; selection of the high-frequency short patterns as features, then the generation of motif candidates from the selected features and finally the motif extraction. The train and test data set of the extracted motif was generated by using sum square error. Then the classification performance was tested with ANFIS which is a kind of the neural-fuzzy system.