EcircPred: Sequence and secondary structural property based computational identification of exonic circular RNAs

Area of Science:

• Bioinformatics
• Molecular Biology
• Genomics

Background:

• Circular RNAs (circRNAs) are stable non-coding RNAs with tissue-specific expression, acting as gene regulators and potential biomarkers for diseases like cancer.
• The biogenesis of circRNAs shares machinery with pre-mRNA splicing, and some circRNAs possess features like open reading frames, making them difficult to distinguish from mRNAs based on sequence alone.
• Existing methods struggle to differentiate exonic circRNAs from mRNAs due to overlapping sequence properties.

Purpose of the Study:

• To develop and validate a computational method for accurately classifying exonic circular RNAs from messenger RNAs.
• To explore the utility of sequence and predicted secondary structure features in discriminating between circRNAs and mRNAs.
• To establish a robust classification model using Artificial Neural Network (ANN) classifiers and a decision voting system.

Main Methods:

• Extraction of sequence-derived features including di-nucleotide index statistics, emission probabilities, and di-nucleotide entropy.
• Development of ANN classifier models based on these extracted features and predicted secondary structures.
• Application of a decision voting strategy to combine outputs from multiple classifiers for final classification.
• Performance evaluation using 10-fold cross-validation.

Main Results:

• The developed classification method achieved an average efficiency of 0.8374, sensitivity of 0.8544, specificity of 0.8203, and Mathews correlation coefficient of 0.6753.
• The study successfully discriminated exonic circular RNAs from mRNAs using sequence and sequence-derived properties.
• This represents the first report of differentiating exonic circRNAs from mRNAs based on sequence and derived features.

Conclusions:

• The proposed computational approach effectively distinguishes exonic circRNAs from mRNAs.
• Sequence and predicted secondary structure features are valuable for differentiating these RNA types.
• This work provides a foundation for further research into circRNA function and application as biomarkers.