PseUdeep: RNA Pseudouridine Site Identification with Deep Learning Algorithm.

Area of Science:

• Computational Biology
• Bioinformatics
• Genomics

Background:

• Pseudouridine (Ψ) is a crucial RNA modification involved in various biological processes.
• Identifying Ψ modification sites is vital for understanding disease mechanisms.
• Traditional laboratory methods for Ψ site identification are costly and time-intensive, necessitating computational approaches.

Purpose of the Study:

• To develop a deep learning framework, PseUdeep, for accurate identification of pseudouridine (Ψ) sites in RNA sequences.
• To enhance the efficiency and reduce the computational cost associated with Ψ site prediction.

Main Methods:

• Employed three RNA sequence feature extraction methods: one-hot encoding, K-tuple nucleotide frequency, and position-specific nucleotide composition.
• Utilized a deep learning architecture involving double convolution, a capsule neural network, and a bidirectional gated recurrent unit with self-attention.
• Applied the framework to predict Ψ sites in *H. sapiens*, *S. cerevisiae*, and *M. musculus*.

Main Results:

• PseUdeep achieved superior prediction accuracy on independent test datasets (S-200 and H-200), with accuracy improvements of 12.38% and 0.68%, respectively.
• The derived feature dimensions were significantly smaller (109, 109, and 119 for the three species) compared to traditional algorithms.
• The model demonstrated superior performance over existing machine learning methods via tenfold cross-validation and independent testing.

Conclusions:

• PseUdeep offers a highly accurate and efficient deep learning approach for pseudouridine site identification.
• The framework provides a valuable computational tool for RNA modification research, aiding in disease mechanism and biological process studies.
• The source code and datasets are publicly available, facilitating further research and application.