Predicting MiRNA-Disease Association by Latent Feature Extraction with Positive Samples

Area of Science:

• Bioinformatics
• Genomics
• Computational Biology

Background:

• MicroRNAs (miRNAs) are crucial in understanding disease etiology and pathogenesis.
• Identifying miRNA-disease associations (MDAs) is vital for medical research.
• Existing methods often use negative samples, leading to prediction inaccuracies.

Purpose of the Study:

• To develop a novel method for predicting potential miRNA-disease associations (MDAs).
• To overcome limitations of existing methods that use negative samples.
• To improve the accuracy of identifying novel MDAs.

Main Methods:

• Latent Feature Extraction for miRNA-Disease Association prediction (LFEMDA).
• Constructing a novel miRNA similarity matrix.
• Integrating disease similarity, known MDAs, and miRNA similarity.
• Utilizing auxiliary variables from miRNA and disease knowledge.

Main Results:

• LFEMDA demonstrates superior performance compared to existing methods on benchmark datasets.
• The method shows effectiveness in identifying associations for both high-association and novel diseases.
• A case study on breast neoplasms validates the method's capacity for uncovering potential MDAs.

Conclusions:

• LFEMDA offers a more accurate approach to predicting miRNA-disease associations by exclusively using positive samples.
• The method enhances the discovery of potential MDAs, contributing to disease etiology research.
• LFEMDA provides a valuable tool for advancing our understanding of miRNA roles in diseases.