A method for miRNA diffusion association prediction using machine learning decoding of multi-level heterogeneous...

Area of Science:

• Biochemistry and Molecular Biology
• Computational Biology
• Genetics

Background:

• MicroRNAs (miRNAs) are vital non-coding RNAs regulating disease processes.
• Accurate prediction of miRNA-disease relationships is essential for diagnostics and therapeutics.
• Existing models face challenges in information retention during prediction.

Purpose of the Study:

• To develop a novel computational approach, MHXGMDA, for predicting miRNA-disease associations.
• To enhance information preservation during the encoding-decoding process for improved accuracy.
• To leverage multi-layer heterogeneous graph structures and machine learning for robust predictions.

Main Methods:

• MHXGMDA utilizes a multi-layer heterogeneous encoder to generate miRNA and disease embeddings.
• Multi-view similarity matrices serve as input for the encoder.
• An XGBoost classifier acts as the machine learning decoder, integrating concatenated layer information.

Main Results:

• MHXGMDA demonstrated superior performance on two benchmark datasets for human miRNA-disease association prediction.
• The method outperformed several leading prediction approaches.
• Performance was validated using Area Under the Receiver Operating Characteristic Curve and Area Under the Precision-Recall Curve metrics.

Conclusions:

• The proposed MHXGMDA method effectively predicts miRNA-disease associations.
• The multi-layer heterogeneous encoder-machine learning decoder structure enhances information preservation and prediction accuracy.
• MHXGMDA offers a promising tool for advancing miRNA-disease association research.