Exploring associations of non-coding RNAs in human diseases via three-matrix factorization with hypergraph-regular...

Area of Science:

• Biomedical Informatics
• Computational Biology
• Genomics

Background:

• Accurate associations between non-coding RNAs (ncRNAs) and diseases are crucial for human biomedical research and treatment.
• Traditional experimental methods for identifying ncRNA-disease associations are time-consuming and expensive.
• Existing computational tools often focus on single ncRNA types or specific diseases.

Purpose of the Study:

• To develop an effective computational predictor for identifying general ncRNA-disease associations.
• To address the limitations of existing methods by considering multiple ncRNA types (circRNAs, miRNAs, lncRNAs).
• To improve the accuracy and efficiency of ncRNA-disease association prediction.

Main Methods:

• Proposed a novel computational method: three-matrix factorization with hypergraph regularization terms (HGRTMF) based on central kernel alignment (CKA).
• Utilized various similarity matrices for circRNAs, miRNAs, and lncRNAs during matrix construction.
• Employed 5-fold cross-validation and leave-one-out cross-validation for performance evaluation.

Main Results:

• Achieved excellent performance across five datasets involving three types of ncRNAs.
• Obtained high Area Under the Curve (AUC) scores, with best scores reaching 0.9832 (5-fold CV) and 0.9836 (LOOCV).
• Demonstrated the method's capability to accurately discover novel ncRNA-disease associations.

Conclusions:

• The proposed CKA-HGRTMF method is a powerful and accurate tool for predicting general ncRNA-disease associations.
• This approach offers a significant advancement over traditional methods and existing computational tools.
• The findings have strong implications for advancing human biomedical research and developing novel disease treatments.