NeRNA

NeRNA generates high-quality negative datasets from known noncoding RNA (ncRNA) sequences to support supervised machine learning for ncRNA classification and identification.

Key Features:

• Negative Data Generation: Transforms known ncRNA sequences and their structural information into octal representations to create negative examples that mimic frameshift mutations without insertions or deletions.
• Versatility Across RNA Types: Validated on microRNA (miRNA), transfer RNA (tRNA), long noncoding RNA (lncRNA), and circular RNA (circRNA) datasets.
• Species-Specific Analysis: Includes species-specific case analysis for miRNA prediction.
• Integration with Machine Learning Models: Evaluated using Decision Tree, Naïve Bayes, Random Forest, Multilayer Perceptron, Convolutional Neural Network (CNN), and Simple feedforward Neural Networks with 1000-fold cross-validation.

Scientific Applications:

• ncRNA classification and identification: Provides high-quality negative controls to improve training and accuracy of supervised classifiers for ncRNA detection.
• Novel sequence identification: Supports discovery of novel noncoding sequences by enabling robust model discrimination between true ncRNAs and negative examples.
• Genomics and transcriptomics: Facilitates analyses of ncRNA functions and interactions by improving model-based inference in genomic and transcriptomic studies.

Methodology:

Transforms sequences and structural information into octal representations to generate negative examples that mimic frameshift mutations without insertions or deletions, and evaluates datasets using Decision Tree, Naïve Bayes, Random Forest, Multilayer Perceptron, Convolutional Neural Network (CNN), and Simple feedforward Neural Networks with 1000-fold cross-validation.