iPseU-NCP

iPseU-NCP predicts pseudouridine (Ψ) sites in RNA sequences using Random Forest models trained on nucleotide chemical property (NCP) features for computational identification of RNA Ψ-modifications.

Key Features:

• Random Forest algorithm: Uses Random Forest classifiers for prediction of pseudouridine sites.
• Nucleotide Chemical Property (NCP) encoding: Represents RNA sequences with NCP-derived features for model input.
• Benchmark training dataset: Model development used the benchmark dataset from Chen et al. (2016).
• Independent evaluation datasets: Performance was evaluated on H. sapiens (H_200) and S. cerevisiae (S_200) datasets.
• Comparative methods: Performance was compared against iPseU-CNN, PseUI, and iRNA-PseU.
• MCC improvements: Reported increases in Matthew's correlation coefficient (MCC) for H. sapiens of ≈20.0%, 55.0%, and 109.0%, and for S. cerevisiae of ≈6.5%, 35.0%, and 150.0% versus the compared methods.
• Accuracy improvements: Reported accuracy increases for H. sapiens of ≈7.0%, 13.0%, and 20.0%, and for S. cerevisiae of ≈2.0%, 9.5%, and 25.0% versus the compared methods.

Scientific Applications:

• Pseudouridine site identification across RNA types: Applicable to rRNA, mRNA, tRNA, and nuclear/nucleolar RNA pseudouridylation mapping.
• Support for biomedical research: Provides computational evidence relevant to studies in drug discovery, gene therapy, and investigations of human disease-associated RNA modifications.

Methodology:

Random Forest models were trained on the Chen et al. (2016) benchmark dataset using NCP-encoded features and evaluated on independent H. sapiens (H_200) and S. cerevisiae (S_200) datasets with performance measured by Matthew's correlation coefficient (MCC) and accuracy against iPseU-CNN, PseUI, and iRNA-PseU.