PsoEL-PseU

PsoEL-PseU predicts pseudouridine sites in RNA sequences by fusing six optimized feature descriptors using binary particle swarm optimization and a parallel fusion ensemble to enhance prediction accuracy and generalization.

Key Features:

• Comprehensive Feature Exploration: Systematically explores various feature descriptors and selects six distinct descriptors relevant to pseudouridine site prediction.
• Optimal Feature Subset Selection: Applies a binary particle swarm optimizer (binary PSO) to identify optimal feature subsets for each descriptor.
• Individual Predictor Training: Trains six individual predictors using the optimized feature subsets derived from the six descriptors.
• Ensemble Learning Strategy: Integrates the six individual predictors via a parallel fusion ensemble to combine model outputs and improve robustness.
• Evaluation Protocols: Evaluates performance using ten-fold cross-validation and independent dataset testing on benchmark datasets.

Scientific Applications:

• Pseudouridine site identification: Identifies pseudouridine sites in RNA sequences across benchmark datasets for studies of RNA pseudouridylation.
• Performance benchmarking and validation: Demonstrates higher accuracy and generalization than existing predictors through ten-fold cross-validation and independent dataset evaluation.

Methodology:

Extracts six feature descriptors, applies binary particle swarm optimization for feature subset selection, trains six individual predictors on optimized subsets, fuses predictions via a parallel ensemble (parallel fusion), and evaluates using ten-fold cross-validation and independent dataset testing.