Multi-dimensional feature recognition model based on capsule network for ubiquitination site prediction.

Area of Science:

• Biochemistry
• Computational Biology
• Proteomics

Background:

• Ubiquitination is a crucial post-translational modification regulating cellular functions.
• Experimental identification of ubiquitination sites is laborious and expensive.
• Existing computational methods, including traditional machine learning and deep learning, have limitations in handling large-scale proteomic data and capturing fine-grained features.

Purpose of the Study:

• To develop an advanced computational model for accurate and efficient prediction of protein ubiquitination sites.
• To overcome the limitations of existing machine learning and deep learning approaches in feature representation and dependency analysis.

Main Methods:

• Proposed a multi-dimensional feature recognition model named MDCapsUbi, utilizing a capsule network architecture.
• Integrated convolution operations and channel attention for coarse-grained feature extraction.
• Employed capsule vectors within the capsule network for fine-grained feature identification and classification of ubiquitination sites.

Main Results:

• MDCapsUbi achieved high performance metrics, including 91.82% accuracy, 91.39% sensitivity, 92.24% specificity, 0.837 MCC, 0.918 F-Score, and 0.97 AUC via ten-fold cross-validation.
• The proposed model demonstrated superior performance compared to existing ubiquitination site prediction technologies.

Conclusions:

• MDCapsUbi offers a powerful and accurate computational approach for protein ubiquitination site prediction.
• The model's architecture effectively captures both coarse-grained and fine-grained features, enhancing prediction accuracy for large-scale proteomic data.