CFCPred: Advancing circRNAs-Encoded Peptides Prediction Through Cluster Purity-Guided Resampling and Fuzzy Voting

Area of Science:

• Genomics
• Computational Biology
• Molecular Biology

Background:

• Circular RNAs (circRNAs) exhibit coding potential through cap-independent translation.
• circRNAs-encoded peptides (circPEPs) are implicated in disease pathogenesis.
• Computational tools for circPEPs prediction are currently underdeveloped.

Purpose of the Study:

• To introduce CFCPred, a novel computational tool for predicting circPEPs.
• To address the challenges of data scarcity and class imbalance in circPEPs prediction.
• To provide a framework for accelerating research on circPEPs and their disease roles.

Main Methods:

• CFCPred integrates cluster purity-guided resampling with fuzzy voting (FV) in a machine learning framework.
• Resampling strategies generate synthetic samples to mitigate class imbalance in training data.
• Fuzzy voting enables adaptive model selection and dynamic weight adjustment for robust consensus classification.

Main Results:

• CFCPred demonstrates strong predictive performance across multiple independent datasets.
• The tool effectively addresses class imbalance and class overlap issues in circPEPs prediction.
• Experimental validation confirms the efficacy of the proposed computational approach.

Conclusions:

• CFCPred offers a robust computational framework for circPEPs prediction.
• The tool can significantly accelerate research into the functional roles of circPEPs in disease.
• This advancement facilitates a deeper understanding of circRNA coding potential and its implications.