SCMFMDA: Predicting microRNA-disease associations based on similarity constrained matrix factorization

Area of Science:

• Biomedical Informatics
• Genomics
• Computational Biology

Background:

• MicroRNAs (miRNAs) are small non-coding RNAs involved in complex biological processes.
• Numerous studies link miRNAs to various human diseases, highlighting their diagnostic and therapeutic potential.
• Accurate prediction of miRNA-disease associations is crucial for understanding disease mechanisms and developing targeted therapies.

Purpose of the Study:

• To propose a novel computational model, Similarity Constrained Matrix Factorization for miRNA-Disease Association Prediction (SCMFMDA), for predicting associations between miRNAs and diseases.
• To effectively integrate diverse miRNA and disease similarity data for enhanced prediction accuracy.

Main Methods:

• Applied a similarity network fusion algorithm to generate integrated disease and miRNA similarity networks.
• Incorporated disease functional similarity, semantic similarity, and Gaussian interaction profile kernel similarity.
• Integrated miRNA functional similarity, sequence similarity, and Gaussian interaction profile kernel similarity.
• Enhanced the traditional Nonnegative Matrix Factorization algorithm with L2 regularization and similarity constraint terms.

Main Results:

• SCMFMDA achieved high prediction performance with AUCs of 0.9675 (global Leave-one-out cross validation) and 0.9447 (five-fold cross validation).
• Case studies on two common human diseases demonstrated the model's practical prediction accuracy.
• A significant portion of the top predicted miRNAs were validated by existing experimental reports.

Conclusions:

• SCMFMDA is an effective computational tool for predicting miRNA-disease associations.
• The integration of multiple similarity measures significantly improves prediction accuracy.
• The model holds promise for advancing research in miRNA-related disease mechanisms and therapeutic strategies.