MvKFN-MDA: Multi-view Kernel Fusion Network for miRNA-disease association prediction

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Identifying microRNA-disease associations is crucial for understanding human diseases.
• Experimental methods for association identification are costly and time-consuming.
• Computational approaches offer an efficient alternative for predicting potential miRNA-disease links.

Purpose of the Study:

• To develop a novel computational method, Multi-view Kernel Fusion Network based prediction method (MvKFN-MDA), for predicting miRNA-disease associations.
• To effectively fuse multiple similarity kernels from diverse data sources using a novel multiple kernel fusion framework.
• To enhance the accuracy and reliability of miRNA-disease association predictions.

Main Methods:

• A Multi-view Kernel Fusion Network (MvKFN) was developed to nonlinearly fuse similarity kernels from various data sources (sequence, functional, semantic, Gaussian profile) for both miRNAs and diseases.
• Integrated similarity kernels for miRNAs and diseases were generated, and feature representations were extracted.
• A neural matrix completion framework was employed for end-to-end learning and association prediction.

Main Results:

• The MvKFN-MDA method demonstrated superior performance compared to existing state-of-the-art methods, achieving higher AUCs in 5-fold cross-validation (5-FCV) and leave-one-out cross-validation (LOOCV).
• Experimental validation confirmed a high proportion of predicted miRNAs for colon cancer (49/50), lymphoma (48/50), and kidney cancer (47/50).
• A case study on breast cancer showed 100% accuracy in predicting top 50 miRNAs, even for a disease with no prior known related miRNAs.

Conclusions:

• The MvKFN-MDA method provides a powerful and accurate computational tool for predicting miRNA-disease associations.
• The novel MvKFN framework effectively integrates heterogeneous data, improving prediction accuracy.
• This approach has significant implications for accelerating the identification of disease-related miRNAs and advancing human disease research.