Biyosekans veri analizinde bilgisayar-tabanlı yaklaşımlar

The structure and function of proteins are closely related. Signaling proteins take part in many biological activities like protein secretion, transportation and more. Signaling proteins are an important topic in drug discovery. For these reasons it is important to classify whether a protein is signaling or not. Machine learning solutions are a good option for this because experimental solutions are time consuming and expensive. The aim of this thesis is to classify signaling proteins using protein encoding schemes and machine learning algorithms. 1867 signaling and 3317 non-signaling proteins were downloaded. Then they were transformed using pseudo amino acid composition (PseAAC) and dipeptide composition to create two datasets. Deep neural network, random forest and support vector machine algorithms were used to classify signaling proteins. For both datasets, the accuracy rate of all models were higher than 0.70. It shows that these models are effective in signal protein classification used with protein encoding schemes.