Protein Deep Learning Classification Using 3D Features
Alaaeldin, Farida; Afify, Yasmine M.; Ismail, Rasha Mohamed; Lotfy Badr, Nagwa;
Abstract
Protein analysis relevance in predicting protein function cannot be ignored. The ability to accurately classify proteins based on their sequences is important for the analysis process. Several feature extraction approaches are used to extract features from protein sequences, either 2D or 3D. The objective of this paper is to investigate the impact of using 3D features in deep learning classification. To achieve this objective, comprehensive experiments were conducted using 3 datasets: two different Phage Virion Proteins databases, as well as a disease dataset. In a nutshell, the Conjoint Triad Method is used to extract 3D features, which are then fed to Convolutional Neural Network. Four popular evaluation metrics were employed to study the classification accuracy of deep learning model using 3D features. Using the proposed technique, promising results are obtained. The maximum training and independent accuracies were achieved, with 86% training accuracy using the disease dataset and 71% independent accuracy using the PVP-Benchmark dataset.
Other data
Title | Protein Deep Learning Classification Using 3D Features | Authors | Alaaeldin, Farida ; Afify, Yasmine M. ; Ismail, Rasha Mohamed; Lotfy Badr, Nagwa | Keywords | CNN, Conjoint Triad Method, Feature Extraction, Classification, Protein | Issue Date | 2021 | Publisher | IEEE | Conference | https://ieeexplore.ieee.org/document/9694247 | ISBN | 978-1-6654-4076-9 | DOI | 10.1109/ICICIS52592.2021.9694247 |
Attached Files
File | Description | Size | Format | Existing users please Login |
---|---|---|---|---|
Protein Deep Learning Classification Using 3D Features.pdf | 657.33 kB | Adobe PDF | Request a copy |
Similar Items from Core Recommender Database
Items in Ain Shams Scholar are protected by copyright, with all rights reserved, unless otherwise indicated.