Benchmarking of computational methods for predicting circRNA-disease associations.

Area of Science:

• Biochemistry
• Bioinformatics
• Genomics

Background:

• Circular RNAs (circRNAs) are increasingly recognized for their roles in human diseases.
• Experimental identification of circRNA-disease associations is time-consuming.
• Computational methods are emerging as efficient alternatives.

Purpose of the Study:

• To systematically categorize existing computational methods for circRNA-disease association prediction.
• To compare the performance of different categories of methods (information propagation, traditional machine learning, deep learning).
• To evaluate method effectiveness using multiple datasets and cross-validation strategies.

Main Methods:

• Categorization of computational methods into information propagation, traditional machine learning, and deep learning.
• Selection and detailed introduction of baseline methods within each category.
• Comparative analysis of 14 representative methods across 5 datasets using 5-fold, 10-fold cross-validation, and de novo experiments.

Main Results:

• Performance evaluation of selected methods across different datasets and validation strategies.
• Comparison of correctly identified circRNA-disease associations for six common cancers.
• Observations on the robustness and characteristics of various computational approaches.

Conclusions:

• Summary of the strengths and weaknesses of different computational methods for circRNA-disease association prediction.
• Identification of key challenges and future research directions in the field.
• Highlighting the potential of computational approaches for advancing circRNA-disease research.