A novel information diffusion method based on network consistency for identifying disease related microRNAs

Area of Science:

• Bioinformatics
• Genomics
• Computational Biology

Background:

• Abnormal microRNA (miRNA) expression is linked to human diseases, making miRNA prediction crucial for diagnosis and treatment.
• Computational methods for inferring miRNA-disease relationships supplement biological experiments, aiding disease prevention and prognosis.

Purpose of the Study:

• To propose a novel information diffusion method based on network consistency (IDNC) for identifying disease-related microRNAs.
• To enhance the accuracy of predicting miRNA-disease associations for improved clinical applications.

Main Methods:

• Reconstructed miRNA and disease networks using miRNA family information, functional similarity, known associations, and semantic disease scores.
• Integrated global network similarity with disease-miRNA association networks using Laplacian scores and network consistency.
• Employed a restarted random walk algorithm to generate prediction scores via stable diffusion spectrum.

Main Results:

• Achieved an Area Under the Curve (AUC) of 0.8814 on a gold standard dataset using Leave-One-Out Cross-Validation (LOOCV).
• Attained an AUC of 0.9512 on a predictive dataset via LOOCV.
• Demonstrated superior accuracy compared to other state-of-the-art methods in identifying disease-related miRNAs.

Conclusions:

• The proposed IDNC method effectively identifies disease-related microRNAs with high accuracy.
• Case studies on breast and colon neoplasms confirm the clinical value and potential of the IDNC method for disease association prediction.