REDInet

REDInet profiles A-to-I RNA editing in human RNA sequencing data using a temporal convolutional network-based deep learning algorithm trained on REDIportal RNA editing sites.

Key Features:

• Deep learning classification: Employs a temporal convolutional network architecture inspired by Google DeepMind's WaveNet to classify editing events with high accuracy.
• Genomic data independence: Classifies editing events using RNAseq nucleotide frequencies of 101-base windows without requiring coupled whole genome sequencing or whole exome sequencing data.
• Large-scale training data: Trained on REDIportal, the largest collection of human A-to-I changes from more than 8000 RNAseq datasets of the Genotype-Tissue Expression project.

Scientific Applications:

• A-to-I RNA editing detection: Identifies adenosine-to-inosine RNA editing events directly from RNA sequencing experiments.
• Noise and artifact reduction: Avoids reliance on empirical filters and cumbersome computational procedures used by current bioinformatics tools to remove background noise, sequencing errors, and artifacts.

Methodology:

REDInet uses a temporal convolutional network-based deep learning algorithm that processes RNAseq nucleotide frequencies from 101-base windows for classification.