A deep learning framework for predicting disease-gene associations with functional modules and graph augmentation.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Understanding gene-disease associations is vital for disease mechanism insights, prevention, and treatment.
• High-throughput biotechnology generates extensive gene-disease data, but prediction methods often miss functional module impacts and data completeness.
• Existing graph representation learning methods for disease-gene association prediction have limitations due to overlooked functional modules and incomplete data.

Purpose of the Study:

• To introduce ModulePred, a novel deep learning framework for predicting disease-gene associations.
• To address limitations in existing methods by incorporating functional modules and improving data representation.

Main Methods:

• ModulePred employs graph augmentation on protein interaction networks using L3 link prediction.
• A heterogeneous module network is constructed integrating disease-gene associations, protein complexes, and augmented protein interactions.
• A novel graph embedding and graph neural network are developed for learning node representations and gene prioritization.

Main Results:

• ModulePred demonstrates superior performance in predicting disease-gene associations compared to existing methods.
• The framework effectively incorporates functional modules and graph augmentation, enhancing prediction accuracy.
• Experimental results validate the effectiveness of the proposed deep learning approach.

Conclusions:

• ModulePred offers an innovative approach to enhance the understanding and prediction of gene-disease relationships.
• Incorporating functional modules and graph augmentation significantly improves disease-gene association prediction.
• This research opens new avenues for leveraging network biology and deep learning in genetic disease research.