NCpred

NCpred predicts and classifies non-coding RNA (ncRNA) sequences to identify long non-coding RNA (lncRNA) elements and Rfam lncRNA families.

Key Features:

• Hybrid Algorithmic Approach: Integrates random forests and logistic regression to enhance classification accuracy using ensemble learning and statistical modeling.
• Discriminative Feature Set: Trained on a well-balanced dataset and uses a discriminative set of features, including a pivotal composite feature named SCORE, to distinguish ncRNA sequences.
• SCORE Feature: Derived from a logistic regression function that combines five attributes—structure, sequence, modularity, structural robustness, and coding potential—to improve characterization of lncRNA elements and identification of Rfam lncRNA families.
• Performance Metrics: Reports overall accuracy of 92.11%, sensitivity of 90.7%, and specificity of 93.5% for ncRNA identification.

Scientific Applications:

• Cross-taxa genomic ncRNA analysis: Applicable across a wide range of organisms, with emphasis on taxa of economic, social, public health, environmental, and agricultural significance.
• Empirical datasets: Applied to bacterial genomes, the Arthrospira (Spirulina) genome, and genomic regions from rice and humans.

Methodology:

Integrates random forests and logistic regression; uses a logistic regression–derived SCORE combining five attributes (structure, sequence, modularity, structural robustness, coding potential); trained on a well-balanced dataset as a classification framework for genome-wide ncRNA analysis, reporting sensitivity >90% for prokaryotic sequences and approximately 77.7% for eukaryotic sequences.