A Semi-Supervised Learning Algorithm for Predicting Four Types MiRNA-Disease Associations by Mutual Information in a...

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• MicroRNA (miRNA) dysregulation is linked to various diseases.
• Accurate identification of miRNA-disease associations is crucial for disease diagnosis and treatment.
• Existing computational methods primarily focus on binary miRNA-disease associations.

Purpose of the Study:

• To develop a novel semi-supervised model for predicting multiple types of miRNA-disease associations.
• To infer underlying miRNA-disease association types using network propagation.
• To overcome limitations of existing binary prediction models.

Main Methods:

• A network-based label propagation algorithm (NLPMMDA) was developed.
• A heterogeneous network was constructed integrating disease semantic similarity, miRNA functional similarity, and Gaussian interaction profile kernel similarity.
• The model utilizes mutual information for inferring associations and does not require negative samples.

Main Results:

• Leave-one-out cross-validation (LOOCV) demonstrated reliable performance across four known miRNA-disease association types.
• Case studies on lung cancer and breast cancer confirmed the model's effectiveness.
• NLPMMDA successfully predicted novel miRNA-disease associations and their specific types.

Conclusions:

• The proposed NLPMMDA model offers a robust approach for predicting multiple types of miRNA-disease associations.
• This method enhances the understanding of miRNA roles in disease pathogenesis.
• NLPMMDA has potential applications in identifying novel disease biomarkers and therapeutic targets.