EPILOGUE: Multi-View Graph Contrastive Learning for Gene Function Prediction.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Accurate gene function prediction is vital for understanding biological systems.
• Existing computational methods struggle with multi-source heterogeneous networks and nonlinear dependencies.
• Contrastive learning offers a promising approach for rich feature representation.

Purpose of the Study:

• To propose EPILOGUE, a multi-view graph contrastive learning framework for improved gene function prediction.
• To leverage graph neural networks and contrastive learning for high-quality gene representations.
• To incorporate protein sequences as node features for comprehensive semantic learning.

Main Methods:

• Developed EPILOGUE, a framework integrating graph neural networks with contrastive learning.
• Utilized multi-view graph learning to capture complex relationships within biological networks.
• Incorporated protein sequences as node features alongside network topology.

Main Results:

• EPILOGUE demonstrated superior performance in gene function prediction compared to nine state-of-the-art methods.
• The framework achieved high effectiveness across six evaluation metrics on yeast and human datasets.
• Generated high-quality, discriminative gene representations for accurate functional annotation.

Conclusions:

• EPILOGUE effectively addresses limitations in current gene function prediction approaches.
• The integration of graph contrastive learning and protein sequence features enhances representation learning.
• The framework offers a robust solution for accurate gene function annotation in biological research.