Current Proceedings on Technology
Yazarlar: Murat Gok, Ahmet Turan Ozcerit, Ayhan Istanbullu
Konular:-
Anahtar Kelimeler:HIV-1 protease specificity,Feature encoding technique,Peptide classification
Özet: Computer aided diagnoses systems have been frequently used in the prediction of HIV-1 protease cleavage site techniques. However, owing to the relatively low performances of encoding schemes and lack of an effective comparison, it is still difficult for researchers to determine which encoding technique is the best. In this paper, we developed a new feature encoding technique to predict HIV-1 protease cleavage sites. We also tested and compared our technique with several encoding schemes from the point of accuracy and area under receiver operating characteristic curve performances. The method we developed utilizes the best specified physicochemical properties of amino acids and each input is encoded accordingly. We employed linear support vector machines and radial basis support vector machines to obtain the performance of each encoding techniques. According to obtained simulation results, our method improves the accuracy rate up to 95.70 % and AUC up to 0.988 when linear support vector machines algorithm is used.