Predicting and classifying functional sites of proteins using support vector machines


3rd International Eurasian Conference onBiological and Chemical Sciences(EurasianBioChem 2020), Ankara, Turkey, 19 - 20 March 2020, pp.532

  • Publication Type: Conference Paper / Summary Text
  • City: Ankara
  • Country: Turkey
  • Page Numbers: pp.532
  • Inonu University Affiliated: Yes


Predicting and classifying functional sites of proteins using support vector machines


Samet Kocabay1*, Hıncal Gökhan Bakır1, Ekrem Atalan1


1 Inonu University, Science and Literature Faculty, Department of Molecular Biology and Genetics, Malatya, Turkey.


*Corresponding author e-mail: 


Protein and other biomolecules perform their interactions and, accordingly, their functions through special sites on their sequence. For this reason, the identification of these special regions defined as motifs is an important bioinformatics application in analyzing the structure and functions of biomolecules. Among the many approaches developed for determining these motifs, approaches that make estimation and classification by learning from experimental data are a current field of study. Support Vector Machines is one of the most common methods among statistical supervised learning methods and has many applications in the field of computational molecular biology.

In this study, functional regions on some selected protein families and groups are classified using experimental data using SVMs and the estimation of functional regions using SVMs is investigated. According to the results obtained in the study carried out in PERL language and by creating functional region word databases-dictionaries, SVMs make successful predictions even in limited datasets and potentially have a wide range of applications. Effects of tuning the parameters such as Kernel design are discussed.

Keywords: Bioinformatics, PERL, regulatory sites, dictionary-based algorithms, SVM