PseU-ST: A new stacked ensemble-learning method for identifying RNA pseudouridine sites

Area of Science:

• Computational Biology
• RNA Modifications
• Bioinformatics

Background:

• Pseudouridine (Ψ) is a prevalent RNA modification vital for numerous biological functions.
• Accurate identification of pseudouridine sites is essential for understanding its roles.
• Experimental methods for Ψ site identification are costly and time-consuming.

Purpose of the Study:

• To develop an efficient computational model for predicting pseudouridine (Ψ) sites in RNA sequences.
• To identify key sequence features that contribute to accurate Ψ site prediction.
• To provide a reliable tool for researchers studying RNA modifications in various species.

Main Methods:

• Proposed the PseU-ST model for Ψ site prediction in *Homo sapiens*, *Saccharomyces cerevisiae*, and *Mus musculus*.
• Evaluated multiple RNA sequence encoding schemes and machine learning algorithms.
• Employed feature selection algorithms (chi-square, incremental feature selection) and a stacking strategy for model optimization.

Main Results:

• PseU-ST demonstrated superior prediction performance compared to existing models.
• Achieved high accuracy scores: 93.64% (H. sapiens), 87.74% (S. cerevisiae), and 89.64% (M. musculus).
• Showcased significant improvements over state-of-the-art methods on benchmark datasets.

Conclusions:

• PseU-ST is a highly competitive and accurate model for identifying RNA pseudouridine sites.
• Position-specific trinucleotide propensity (PSTNPss) and PS3 features are crucial for Ψ site identification.
• The PseU-ST model and associated data are publicly available on GitHub for broader research use.