Related Experiment Video
Updated: Jun 7, 2025
Wide-Field, Real-Time Imaging of Local and Systemic Wound Signals in Arabidopsis
Published on: June 4, 2021
WGAN-GP_Glu: A semi-supervised model based on double generator-Wasserstein GAN with gradient penalty algorithm for
1Information Science and Technology, Dalian Maritime University, Dalian, Liaoning, China; The School of Artificial Intelligence and Computer Science, Jiangnan University, Wuxi, China; Neusoft Education Technology Group, Dalian, China; Key Laboratory of Symbolic Computation and Knowledge Engineering of Ministry of Education, Jilin University, Changchun, 130012, China.
This study introduces WGAN-GP_Glu, a novel semi-supervised learning algorithm for identifying glutarylation sites. It effectively addresses class imbalance, improving accuracy in predicting these crucial post-translational modifications.
Area of Science:
- Biochemistry and Molecular Biology
- Computational Biology
- Bioinformatics
Background:
- Glutarylation is a vital post-translational modification influencing cellular functions.
- Existing computational methods for glutarylation site identification face challenges with data noise and class imbalance.
- Accurate identification of non-glutarylation sites is critical for robust computational models.
Purpose of the Study:
- To develop a novel semi-supervised learning algorithm, WGAN-GP_Glu, for accurate glutarylation site identification.
- To address the challenge of class imbalance in identifying non-glutarylation lysine sites.
- To improve the reliability and performance of computational tools for predicting glutarylation.
Main Methods:
- Proposed WGAN-GP_Glu, a multi-module framework including reliable negative sample selection, deep feature extraction, and prediction.
- Developed ReliableWGAN-GP, an improved Wasserstein GAN with Gradient Penalty, for selecting reliable non-glutarylation samples.
- Employed Convolutional Neural Network (CNN) and Bidirectional Long Short-Term Memory (BiLSTM) with attention for deep feature extraction.
Main Results:
- WGAN-GP_Glu achieved high performance on an independent test set: 90.58% sensitivity, 95.82% specificity, 94.44% accuracy, and 0.8645 Matthew correlation coefficient.
- The ReliableWGAN-GP algorithm effectively selected reliable negative samples, mitigating data noise.
- The proposed method outperformed existing approaches for glutarylation site prediction.
Conclusions:
- WGAN-GP_Glu is a powerful and effective tool for identifying glutarylation sites.
- The ReliableWGAN-GP algorithm demonstrates efficacy in selecting reliable negative samples, crucial for addressing class imbalance.
- The findings contribute to advancing computational methods in post-translational modification analysis.
