A viral protein identifying framework based on temporal convolutional network.

Area of Science:

• Biopharmaceuticals
• Computational Biology
• Drug Discovery

Background:

• Viral protein identification from amino acid sequences is crucial for drug design.
• Traditional methods struggle with data imbalance and long computation times.
• Accurate viral protein identification is a fundamental challenge in biopharmaceuticals.

Purpose of the Study:

• To develop an efficient deep learning framework for identifying viral proteins.
• To overcome data imbalance and reduce computational costs in viral protein prediction.
• To enhance the accuracy and speed of viral protein identification for drug design.

Main Methods:

• Utilized Temporal Convolutional Network (TCN) for efficient feature extraction, replacing Recurrent Neural Networks (RNN).
• Implemented a cost-sensitive loss function within TCN to handle data imbalance.
• Integrated misclassification costs into Gradient Boosting Decision Tree (GBDT) weight updates to address imbalanced training data.

Main Results:

• The proposed deep learning framework significantly outperforms traditional methods for imbalanced data.
• Demonstrated a substantial reduction in computation time compared to existing approaches.
• Achieved slight performance enhancements in viral protein identification accuracy.

Conclusions:

• The novel deep learning framework offers an efficient and accurate solution for viral protein identification.
• The TCN and GBDT integration effectively tackles data imbalance and computational challenges.
• This approach advances drug design by improving the speed and reliability of viral protein analysis.