REDInet: a temporal convolutional network-based classifier for A-to-I RNA editing detection harnessing million known...

Area of Science:

• Bioinformatics
• Genomics
• Computational Biology

Background:

• A-to-I RNA editing detection is challenging.
• Current bioinformatics tools require empirical filters and extensive sequencing data, often involving complex computational steps.
• Existing methods struggle with background noise, sequencing errors, and artifacts.

Purpose of the Study:

• To present REDInet, a novel deep learning algorithm for profiling RNA editing in human RNA sequencing (RNAseq) data.
• To develop a method that overcomes the limitations of current RNA editing detection tools.
• To accurately classify RNA editing events without relying on coupled genomic data.

Main Methods:

• Developed REDInet, a temporal convolutional network-based deep learning algorithm.
• Trained REDInet on the REDIportal database, comprising over 8000 RNAseq datasets from the genotype-tissue expression project.
• Utilized RNAseq nucleotide frequencies within 101-base windows for classification.

Main Results:

• REDInet demonstrates high accuracy in classifying RNA editing events.
• The algorithm effectively profiles RNA editing directly from RNAseq data.
• No requirement for whole genome sequencing or whole exome sequencing data is needed.

Conclusions:

• REDInet offers an efficient and accurate method for A-to-I RNA editing detection.
• The deep learning approach simplifies the RNA editing profiling process.
• REDInet provides a valuable tool for analyzing RNA editing in human transcriptomes.