Leveraging explainable multi-scale features for fine-grained circRNA-miRNA interaction prediction.

Area of Science:

• Bioinformatics
• Computational Biology
• Molecular Biology

Background:

• Circular RNAs (circRNAs) and microRNAs (miRNAs) interactions are crucial in biological processes and diseases.
• Computational methods are vital for predicting these molecular interactions.
• Existing methods lack precise interpretability and effective molecular representation.

Purpose of the Study:

• To develop a novel computational method for predicting circRNA-miRNA interactions (CMI).
• To address limitations in model interpretability and molecular representation in current prediction methods.

Main Methods:

• Proposed MFERL method utilizing multi-scale representation learning.
• Employed explainable fine-grained modeling for CMI prediction.
• Integrated dual-convolution attention and contrastive learning for feature enhancement.

Main Results:

• MFERL achieves robust generalization, robustness, and interpretability.
• The method outperforms existing state-of-the-art models in predicting CMI.
• Manifold-based analysis confirmed detailed model performance.

Conclusions:

• MFERL provides a promising approach for understanding CMI.
• The method offers enhanced accuracy and interpretability in CMI prediction.
• MFERL advances the study of molecular interactions in disease contexts.