CCXGB: Centroid-based features enhancement using Convolutional Neural Network combined with XGB classifier for Protein-Protein interaction prediction
The facts allied to the interaction of proteins are crucial due to their significance in numerous biological and cellular activities. With the accessibility of huge protein-protein interaction datasets, the openings to attain efficient deep network models in prediction applications also has been amp...
Gespeichert in:
Veröffentlicht in: | International journal of information technology (Singapore. Online) 2024, Vol.16 (1), p.393-401 |
---|---|
Hauptverfasser: | , |
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | The facts allied to the interaction of proteins are crucial due to their significance in numerous biological and cellular activities. With the accessibility of huge protein-protein interaction datasets, the openings to attain efficient deep network models in prediction applications also has been amplified. Despite deep networks, Convolutional Neural Network (CNN) is applied to visual imagery analysis and classification, hence only a few mechanisms are using CNN in protein-protein interaction (PPI) prediction. This article proposed a sequence-based PPI prediction method CCXGB, which is evolving the one-dimensional convolution layer (Conv1D)’s performance in prediction by integrating it with the Centroid-based feature extraction method and eXtreme Gradient Boosting (XGB). The CCXGB evolves the performance of the PPI prediction model by feeding centroid-based raw features in the CNN module to extract high-level protein features which are further involved in the classification function of XGB. CCXGB model can predict PPI with 0.999, and 0.996 Area Under Curve (AUC) scores and 99.38%, and 97.98% average accuracy scores for Human, and S.
cerevisiae
datasets respectively. The significance of each module of the proposed model is proved through different comparing approaches. Correspondingly, the CCXGB demonstrated superior performance when compared with existing PPI prediction models. |
---|---|
ISSN: | 2511-2104 2511-2112 |
DOI: | 10.1007/s41870-023-01577-0 |