rnacon

rnacon predicts and classifies non-coding RNAs by discriminating coding from non-coding transcripts and assigning ncRNAs to structural classes using machine learning models.

Key Features:

• Coding vs Non-Coding Discrimination: Applies a Support Vector Machine (SVM) using tri-nucleotide composition as the sole feature, achieving a Matthews Correlation Coefficient (MCC) of 0.98.
• ncRNA Class Classification: Classifies non-coding RNAs into 18 classes using graph properties derived from predicted ncRNA secondary structures and a Random Forest model.
• Comparative Performance: Outperforms BayeNet, NaiveBayes, MultilayerPerceptron, IBk, libSVM, SMO, and GraPPLE, achieving overall sensitivity of 0.43 and MCC of 0.40.

Scientific Applications:

• ncRNA Functional Genomics: Enables accurate identification and subclassification of non-coding RNAs to support studies of regulatory RNA functions.

Methodology:

rnacon uses an SVM trained on tri-nucleotide composition to distinguish coding from non-coding transcripts and applies a Random Forest classifier based on graph features of predicted ncRNA structures to assign sequences to 18 ncRNA classes.