CIRCScan

CIRCScan predicts circular RNA (circRNA) expression across cell lines by integrating sequence and epigenetic features to model presence/absence and quantitative expression levels.

Key Features:

• Feature integration: Integrates sequence features with epigenetic features for circRNA modeling.
• Machine learning models: Employs machine learning algorithms to construct predictive models for circRNA expression status and levels.
• Status prediction performance: Status models achieve ROC area under the curve values of 0.89–0.92 with false-positive rates between 0.17 and 0.25.
• Expression-level prediction performance: Expression-level models report normalized root-mean-square errors of 0.28–0.30, Pearson's r > 0.4, and Spearman's ρ of 0.33–0.46 across tested cell lines.
• Validation data: Predictions were validated using RNA sequencing data.
• Epigenetic driver identification: Identifies the histone modification H3K79me2 as consistently highly ranked for predicting both circRNA expression status and levels.
• Epigenetic enrichment: Reports enrichment of H3K79me2 within intron regions flanking circRNAs across nine additional cell lines.
• Assembler component: Includes an assembler component for circRNA analysis.

Scientific Applications:

• Expression status prediction: Predicts presence or absence of specific circRNAs in different cellular contexts.
• Quantitative expression estimation: Estimates circRNA expression levels across cell lines.
• Epigenetic regulation studies: Enables systematic investigation of how epigenetic modifications, including H3K79me2, influence circRNA expression.
• Non-coding RNA regulatory analysis: Supports exploration of regulatory mechanisms of non-coding RNAs via integration of sequence and epigenetic data.

Methodology:

Integrates sequence and epigenetic features and applies machine learning algorithms to predict circRNA presence/absence and expression levels, with validation using RNA sequencing data and an included assembler component.