Adaptive multi-source multi-view latent feature learning for inferring potential disease-associated miRNAs.

Area of Science:

• Biochemistry
• Genomics
• Computational Biology

Background:

• MicroRNAs (miRNAs) are critical in biological processes, and their dysregulation is linked to cancer.
• In silico identification of disease-associated miRNAs offers a cost-effective approach for biomarker discovery.
• Existing computational methods often rely on single data representations and struggle with novel miRNA-disease associations.

Purpose of the Study:

• To develop a novel computational method for inferring potential disease-associated miRNAs.
• To address limitations of existing methods by integrating multi-source and multi-view data.
• To discover unobserved miRNA-disease associations for new miRNAs or diseases.

Main Methods:

• Adaptive multi-source multi-view latent feature learning (M2LFL) framework.
• Utilizing multiple data sources to capture latent features of miRNA and disease spaces.
• Projecting multi-modal latent features into a common subspace with adaptive joint graph regularization and Lp,q-norms for sparsity and interpretability.

Main Results:

• Demonstrated superior performance in screening reliable candidate disease-associated miRNAs.
• Successfully identified unobserved miRNA-disease associations.
• Validated the method's effectiveness in discovering diagnostic biomarkers.

Conclusions:

• M2LFL is an efficient computational tool for identifying potential disease-associated miRNAs.
• The method enhances biomarker discovery for disease diagnosis and treatment.
• M2LFL provides a robust approach for uncovering novel miRNA-disease relationships.