Inferring the Disease-Associated miRNAs Based on Network Representation Learning and Convolutional Neural Networks

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Identifying disease-associated microRNAs (miRNAs) is crucial for understanding disease mechanisms.
• Existing methods often use shallow models, failing to capture intricate relationships within miRNA-disease networks.

Purpose of the Study:

• To develop a novel prediction method, CNNMDA, for identifying disease-associated miRNAs.
• To deeply integrate miRNA/disease similarities, associations, and network representations using deep learning.

Main Methods:

• Proposed CNNMDA, a framework combining network representation learning and convolutional neural networks.
• Integrated diverse biological data for embedding layers and used non-negative matrix factorization for low-dimensional representations.
• Employed convolutional modules to learn complex relationships in miRNA-disease networks.

Main Results:

• CNNMDA demonstrated superior performance in predicting disease miRNAs compared to state-of-the-art methods.
• Experimental results from cross-validation validated the method's effectiveness.
• Case studies on lung, breast, and pancreatic neoplasms highlighted CNNMDA's ability to discover potential disease miRNAs.

Conclusions:

• CNNMDA offers a powerful deep learning approach for disease miRNA prediction.
• The method effectively captures complex, non-linear relationships in biological networks.
• CNNMDA shows significant potential for advancing etiological and pathogenetic research.