Predicting miRNA-Disease Association Based on Improved Graph Regression.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• MicroRNAs (miRNAs) are increasingly linked to complex diseases and biological processes.
• Accurate models are crucial for inferring miRNA-disease associations.
• Existing biological datasets contain inherent noise that can affect model accuracy.

Purpose of the Study:

• To present an improved model, IGRMDA, for predicting potential miRNA-disease associations.
• To enhance the accuracy of miRNA and disease similarity data by reducing noise.
• To leverage latent spaces for more effective association prediction.

Main Methods:

• Utilized matrix decomposition to process miRNA functional and disease semantic similarity data.
• Combined processed similarity data with existing information to generate final similarity datasets.
• Applied an improved graph regression algorithm within latent spaces (miRNA-disease association, miRNA similarity, disease similarity).
• Incorporated Non-negative Matrix Factorization and Partial Least Squares for attribute extraction.

Main Results:

• The IGRMDA model demonstrated effectiveness in predicting potential miRNA-disease associations.
• Cross-validation experiments confirmed the model's predictive power.
• Case studies validated the model's ability to identify novel miRNA-disease relationships.

Conclusions:

• IGRMDA provides a robust method for inferring miRNA-disease associations.
• The approach effectively handles noise in biological data for improved predictions.
• This model can aid in understanding the complex roles of miRNAs in human diseases.