Data Integration Using Tensor Decomposition for the Prediction of miRNA-Disease Associations

Area of Science:

• Biomedical Informatics
• Computational Biology
• Genomics

Background:

• MicroRNA (miRNA) dysfunction is linked to various diseases through target gene interactions.
• Identifying disease-associated miRNAs is crucial for effective disease prevention and treatment strategies.
• Biological mechanisms underlying miRNA-disease associations are complex, necessitating integrated computational approaches.

Purpose of the Study:

• To propose a novel computational method, TDMDA, for identifying pathogenic microRNAs (miRNAs) associated with diseases.
• To integrate multi-type biological data for enhanced accuracy in miRNA-disease association prediction.
• To leverage tensor decomposition for a comprehensive analysis of complex biological relationships.

Main Methods:

• Constructed a three-order association tensor representing miRNA-disease, miRNA-gene, and gene-disease associations.
• Applied a tensor decomposition method incorporating auxiliary information (biological similarity and adjacency).
• Utilized the reconstructed tensor to predict novel miRNA-disease associations.

Main Results:

• The TDMDA method demonstrated competitive performance compared to existing advanced methods.
• 5-fold cross-validation confirmed the efficacy of the proposed approach.
• The integration of multi-type data via tensor decomposition proved effective for identifying disease-related miRNAs.

Conclusions:

• TDMDA is a robust and competitive method for predicting miRNA-disease associations.
• The tensor decomposition approach offers a powerful framework for integrating complex biological data.
• Accurate identification of disease-related miRNAs using TDMDA can significantly contribute to disease diagnostics and therapeutics.