Improved Prediction of miRNA-Disease Associations Based on Matrix Completion with Network Regularization

Area of Science:

• Biomedical Informatics
• Genomics
• Computational Biology

Background:

• MicroRNAs (miRNAs) are crucial regulators in biological processes and disease pathogenesis.
• Identifying miRNA-disease associations offers insights into disease mechanisms and potential prognostic markers.
• Wet experiments for miRNA-disease association discovery are time-consuming and costly.

Purpose of the Study:

• To develop a computational model for predicting novel miRNA-disease associations.
• To address the limitations of existing models that are biased towards known associations and perform poorly for rare miRNAs and uncommon diseases.
• To provide a cost-effective alternative to experimental methods for discovering miRNA-disease links.

Main Methods:

• A general framework named improved prediction of miRNA-disease associations (IMDN) was developed.
• The framework utilizes matrix completion with network regularization.
• It incorporates miRNA networks as implicit feedback to predict associations based on neighboring relationships.

Main Results:

• IMDN achieved high performance in predicting miRNA-disease associations.
• Reliable area under the receiver operating characteristic (ROC) area under the curve (AUC) values of 0.9162 (global LOOCV) and 0.8965 (local LOOCV) were obtained.
• Case studies validated known associations and identified novel potential miRNA-disease relationships.

Conclusions:

• IMDN is an effective computational approach for discovering potential miRNA-disease associations.
• The method demonstrates robustness in handling data scarcity, particularly for rare miRNAs and uncommon diseases.
• IMDN offers a valuable tool for advancing our understanding of miRNA roles in human diseases and identifying new biomarkers.