COPPER

COPPER predicts plant virus-derived small interfering RNAs (vsiRNAs) using deep learning to identify RNA sequences involved in plant antiviral defense.

Key Features:

• Deep Learning-Based Stacking Ensemble: COPPER employs a stacking ensemble integrating multiple deep learning architectures to improve vsiRNA prediction accuracy.
• Sequence Feature Generation: COPPER uses word2vec and fastText models to generate sequence features that capture complex patterns in RNA sequences.
• Convolutional Neural Network (CNN): The CNN component captures local dependencies in sequence data.
• Multiscale Residual Network: The multiscale residual network extracts features at multiple scales to enhance representation.
• Bidirectional Long Short-Term Memory (BiLSTM) with Self-Attention: The BiLSTM with self-attention models long-range dependencies and focuses on relevant sequence regions to improve interpretability.

Scientific Applications:

• Understanding Antiviral Defense Mechanisms: Identification of vsiRNAs enables molecular-level investigation of plant RNA interference responses to viral infection.
• Development of Antiviral Plants: Predicted vsiRNAs can be used to identify target pathways for engineering plants with enhanced virus resistance.
• Comparative Genomics and Functional Studies: COPPER facilitates comparative analyses across plant species or strains to study the evolution and function of RNA interference mechanisms.

Methodology:

COPPER's methodology comprises extensive benchmarking across various sequence homology thresholds and ablation studies, and comparative evaluation against PVsiRNAPred on independent test datasets demonstrating improved accuracy.