StackCirRNAPred: computational classification of long circRNA from other lncRNA based on stacking strategy

Area of Science:

• Bioinformatics
• Genomics
• Molecular Biology

Background:

• Circular RNAs (circRNAs) are crucial regulators of gene expression and implicated in disease.
• Existing computational tools for circRNA prediction have limitations due to single-classifier approaches.
• Improved methods are needed to accurately identify circRNAs from other long non-coding RNAs (lncRNAs).

Purpose of the Study:

• To develop a novel computational tool, StackCirRNAPred, for enhanced circRNA prediction.
• To improve the accuracy and robustness of distinguishing long circRNAs from other lncRNAs.
• To integrate diverse features and multiple classifiers for superior predictive performance.

Main Methods:

• Extraction of diverse features: nucleic acid composition, sequence spatial features, physicochemical properties, and repeat elements.
• Application of a stacking strategy to integrate Random Forest (RF), LightGBM, and XGBoost classifiers.
• Validation of StackCirRNAPred on human and mouse datasets.

Main Results:

• StackCirRNAPred demonstrated superior performance compared to existing tools.
• Achieved high metrics: precision (0.843), accuracy (0.833), F1 (0.831), recall (0.819), and MCC (0.666) on human data.
• Showed consistent high performance on mouse data: precision (0.837), accuracy (0.839), F1 (0.839), recall (0.841), and MCC (0.677).

Conclusions:

• StackCirRNAPred effectively distinguishes long circRNAs from other lncRNAs using a stacking strategy.
• The tool's validity and robustness were confirmed through rigorous testing.
• StackCirRNAPred offers a valuable addition to existing circRNA prediction methods for downstream research.