Improving the Prediction Performance of Protein Protein Interaction Sites Using Xgboost and Optimization


Abstract views: 32 / PDF downloads: 63

Authors

  • Yunus Emre GÖKTEPE Necmettin Erbakan University

DOI:

https://doi.org/10.59287/icaens.1023

Keywords:

Protein-Protein Interaction Sites, Extreme Gradient Boosting, Hyperparameter Optimization

Abstract

Since proteins have important tasks in all processes within the cell they are vital elements for all living organisms. They are significant in regulating most of the biological processes which occur in a cell. They are widely researched to comprehend roles of them and to assist drug design studies. In these tasks, they usually work by interacting with other proteins, not alone. Thus, predicting protein-protein interactions and protein protein interaction sites is an important problem in bioinformatics. There are a number of computational methods developed for this prediction task. DeepPPISP-XGB is one of these methods and produces promising predicting results. In this study, we proposed an optimization process in order to improve prediction results of this method. This optimization process produced 1.5% better AUROC value and 3.3% better AUPRC value compared to DeepPPISP-XGB method.

Author Biography

Yunus Emre GÖKTEPE, Necmettin Erbakan University

Department of Computer Engineering, Fac. Eng. Arch.  Konya, Turkey

Downloads

Published

2023-07-20

How to Cite

GÖKTEPE, Y. E. (2023). Improving the Prediction Performance of Protein Protein Interaction Sites Using Xgboost and Optimization. International Conference on Applied Engineering and Natural Sciences, 1(1), 365–369. https://doi.org/10.59287/icaens.1023