CNCTDiscriminator: coding and noncoding transcript discriminator - an excursion through hypothesis learning and...

Area of Science:

• Computational Biology
• Bioinformatics
• Genomics

Background:

• Differentiating coding and noncoding transcripts is crucial for understanding gene regulation.
• Traditional methods often rely on limited feature sets.
• Next-Generation Sequencing (RNA-seq) provides extensive transcriptomic data, enabling new classification approaches.

Purpose of the Study:

• To develop and evaluate CNCTDiscriminator, a system for coding and noncoding transcript discrimination.
• To integrate diverse transcript features for improved classification accuracy.
• To compare CNCTDiscriminator's performance against established methods.

Main Methods:

• Utilized open reading frame (ORF) statistics, base composition, and predicted secondary structure properties.
• Integrated RNA-sequencing (RNA-seq) expression profiles as an additional feature dimension.
• Employed hypothesis learning and feature-specific ensemble learning for feature integration.

Main Results:

• CNCTDiscriminator trained with composition and ORF features achieved 83.86% precision and 82.01% recall, outperforming CPC, CPAT, and PORTRAIT on a benchmark dataset.
• The ensemble-trained CNCTDiscriminator demonstrated comparable performance with 89.85% precision and 71.08% recall.
• Feature integration proved effective in enhancing transcript classification.

Conclusions:

• CNCTDiscriminator offers a robust approach to coding and noncoding transcript discrimination.
• Integrating multiple transcript features, including expression profiles, significantly improves classification.
• The proposed system provides a valuable tool for transcriptomic data analysis.