Definer: A computational method for accurate identification of RNA pseudouridine sites based on deep learning

Area of Science:

• Molecular Biology
• Bioinformatics
• Genomics

Background:

• Pseudouridine is a crucial RNA modification found in non-coding RNAs, impacting gene expression, RNA stability, and disease.
• Accurate identification of pseudouridine sites is vital for understanding its functional mechanisms.
• Traditional experimental methods are insufficient for the scale of modern genomics data.

Purpose of the Study:

• To develop a computational method for accurate identification of RNA pseudouridine sites.
• To apply deep learning for predicting pseudouridine modification sites in Homo sapiens, Saccharomyces cerevisiae, and Mus musculus.

Main Methods:

• A deep learning model named Definer was developed.
• Definer utilizes two sequence coding schemes: Nucleotide Composition Profile (NCP) and One-hot encoding.
• The model integrates Convolutional Neural Networks (CNN), Gated Recurrent Units (GRU), and Attention mechanisms.

Main Results:

• Definer demonstrated superior performance compared to existing methods on a benchmark dataset across three species.
• 10-fold cross-validation confirmed the method's robustness.
• Independent testing on Homo sapiens and Saccharomyces cerevisiae datasets validated Definer's predictive accuracy.

Conclusions:

• Definer provides an accurate and efficient computational approach for identifying RNA pseudouridine modification sites.
• The method is applicable across different species, aiding in the functional analysis of RNA modifications.