MetaDegron: multimodal feature-integrated protein language model for predicting E3 ligase targeted degrons

Area of Science:

• Biochemistry
• Molecular Biology
• Computational Biology

Background:

• Protein degradation via the ubiquitin proteasome system is vital for cellular processes.
• E3 ligases and degrons mediate targeted protein destruction, maintaining cellular homeostasis.
• Limited knowledge exists on identified degron instances and their characteristics.

Purpose of the Study:

• To develop a novel deep-learning framework, MetaDegron, for predicting E3 ligase targeted degrons.
• To integrate protein language models and comprehensive featurization for enhanced prediction accuracy.
• To provide functional annotations and visualizations of degron features.

Main Methods:

• Developed MetaDegron, a deep-learning framework.
• Integrated protein language models and advanced featurization strategies.
• Evaluated performance on benchmark datasets and compared with existing methods like Degpred.

Main Results:

• MetaDegron demonstrates superior performance in predicting E3 ligase targeted degrons compared to existing methods.
• The framework enables batch prediction for 21 E3 ligases.
• Provided functional annotations and visualizations of degron structural and physicochemical features.

Conclusions:

• MetaDegron is a powerful tool for identifying and characterizing targeted degrons.
• The tool aids in elucidating protein homeostasis regulation, cancer research, and drug development.
• MetaDegron is freely available for research and clinical applications.