RODAN

RODAN performs RNA basecalling from raw nanopore sequencing signals using a fully convolutional deep learning architecture.

Key Features:

• Fully Convolutional Neural Network: Applies convolutional neural networks (CNNs) to process raw electrical signals from nanopore RNA sequencing without intermediate handcrafted signal processing.
• Enhanced Basecalling Accuracy: Achieves improved performance compared to existing Oxford Nanopore Technologies RNA basecallers, increasing read accuracy for RNA sequencing data.

Scientific Applications:

• RNA Sequencing and Transcriptomic Analysis: Improves basecalling accuracy for transcriptomics, viral genomics, and functional genomics studies involving RNA viruses and transcript variant characterization.

Methodology:

RODAN directly inputs raw nanopore electrical signals into a fully convolutional neural network that learns sequence-dependent signal patterns, models noise and variability inherent to RNA sequencing, and outputs nucleotide sequences through end-to-end deep learning inference.