CPPred: coding potential prediction based on the global description of RNA sequence.

Area of Science:

• Molecular Biology
• Bioinformatics

Background:

• Next-generation sequencing generates numerous novel RNA transcripts, necessitating accurate classification of coding RNAs (coRNAs) and non-coding RNAs (ncRNAs).
• Existing methods like CPAT, CPC2, and PLEK show limitations in distinguishing small coRNAs from small ncRNAs.

Purpose of the Study:

• To develop a superior computational approach for predicting coding potential in RNA sequences.
• To improve the accuracy of differentiating between small coding and non-coding RNAs.

Main Methods:

• Development of CPPred (coding potential prediction), an SVM classifier utilizing multiple sequence features.
• Incorporation of novel RNA features, including global description encoding (T2, C0, GC), into the prediction model.

Main Results:

• CPPred demonstrates enhanced accuracy in distinguishing coRNAs from ncRNAs and small coRNAs from small ncRNAs compared to state-of-the-art methods.
• Application of CPPred to a dataset of proposed novel human coding RNAs revealed that 91.1% were misclassified, identifying them as ncRNAs.
• Global description features (T2, C0, GC) were identified as critical for accurate coding potential prediction.

Conclusions:

• CPPred offers a more robust tool for RNA classification, particularly for small RNA molecules.
• The findings challenge previous assertions of novel coding RNA discoveries, highlighting the prevalence of ncRNAs in such datasets.
• Global RNA encoding features are significant determinants of coding potential.