SSCMDA: spy and super cluster strategy for MiRNA-disease association prediction

Area of Science:

• Computational biology
• Genomics
• Biomedical informatics

Background:

• Identifying microRNA-disease associations is crucial for clinical medicine but experimentally expensive and time-consuming.
• Computational models are increasingly used to predict potential microRNA-disease associations.
• Existing models face challenges with inaccurate predictions due to mixed unknown miRNA-disease pairs.

Purpose of the Study:

• To develop an effective computational model for predicting microRNA-disease associations.
• To address the limitations of existing models in handling unknown miRNA-disease pairs.
• To improve the accuracy and robustness of miRNA-disease association predictions.

Main Methods:

• Proposed the Spy and Super Cluster strategy for MiRNA-Disease Association prediction (SSCMDA).
• Integrated disease similarity and miRNA similarity.
• Employed a 'spy strategy' to identify reliable negative samples from unknown pairs.
• Utilized a 'super-cluster strategy' to maximize positive samples and overcome training data shortages.

Main Results:

• Achieved high AUC values in cross-validation: 0.9007 (global LOOCV), 0.8747 (local LOOCV), and 0.8806+/-0.0025 (5-fold CV).
• Demonstrated significant improvement over previous models.
• Case studies confirmed prediction robustness and effectiveness for new diseases.
• A large proportion of predicted miRNA-disease associations were experimentally verified.

Conclusions:

• SSCMDA offers a significant advancement in predicting microRNA-disease associations.
• The model's strategies effectively handle challenges with unknown and limited training data.
• SSCMDA shows strong potential for clinical applications and discovering novel miRNA-disease links.