Tensor decomposition with relational constraints for predicting multiple types of microRNA-disease associations

Area of Science:

• Genomics
• Computational Biology
• Biomedical Informatics

Background:

• MicroRNAs (miRNAs) are critical regulators in biological processes and human diseases.
• Current computational methods primarily predict binary miRNA-disease associations, overlooking diverse roles.
• Understanding varied miRNA roles, like genetic variants in leukemia or circulating biomarkers for cancer, is crucial.

Purpose of the Study:

• To develop a computational method for predicting multi-type miRNA-disease associations, moving beyond binary predictions.
• To represent miRNA-disease-type interactions as a tensor for advanced analysis.
• To introduce and evaluate tensor decomposition methods for this task.

Main Methods:

• Representing miRNA-disease-type triples as a tensor.
• Applying tensor decomposition techniques to predict multi-type associations.
• Developing Tensor Decomposition with Relational Constraints (TDRC) incorporating biological features.

Main Results:

• Tensor decomposition methods significantly improve prediction performance over existing baselines on HMDD v2.0 and v3.2 datasets (up to 38% in Top-1F1).
• The proposed TDRC method demonstrates superior performance and efficiency compared to other tensor decomposition methods.
• Incorporating biological features as relational constraints enhances prediction accuracy.

Conclusions:

• Tensor decomposition is a powerful approach for predicting multi-type miRNA-disease associations.
• TDRC offers an effective and efficient computational strategy for understanding complex miRNA-disease relationships.
• This work advances the field by providing a more nuanced prediction of miRNA involvement in diseases.