Deepred-Mt: Deep representation learning for predicting C-to-U RNA editing in plant mitochondria

Area of Science:

• Molecular Biology
• Bioinformatics
• Genomics

Background:

• Cytidine-to-uridine (C-to-U) RNA editing is crucial for land plant mitochondrial function.
• Computational prediction of RNA editing sites is vital for genetic engineering and RNA regulation.

Purpose of the Study:

• To develop a precise computational tool for predicting C-to-U RNA editing events in plant mitochondria.
• To overcome limitations of existing prediction methods based on sequence homology.

Main Methods:

• A deep convolutional neural network, Deepred-Mt, was developed to predict C-to-U editing.
• The model utilizes the sequence context of 40 flanking nucleotides around a cytidine.
• Optimization involved editing extent information, data augmentation, and a large dataset from 21 plant mitochondrial genomes.

Main Results:

• Deepred-Mt demonstrated superior predictive performance compared to homology-based methods, achieving higher average precision and F1 scores.
• The model successfully identified known RNA editing sequence motifs.
• Analysis suggests local RNA structure influences editing site regulation.

Conclusions:

• Deepred-Mt is an effective tool for predicting C-to-U RNA editing in plant mitochondria.
• The findings enhance our understanding of RNA editing mechanisms and their regulatory factors.