nanoDoc

nanoDoc detects post-transcriptional modifications (PTMs) in RNA from raw Nanopore single-molecule direct RNA sequencing (DRS) reads by analyzing deviations in electric current signals to identify modification-associated anomalies.

Key Features:

• Deep Learning Approach: Employs a convolutional neural network to extract features from Nanopore current signals and uses Deep One-Class Classification to detect PTMs as anomalous signals.
• High Accuracy: Demonstrates an area under the curve (AUC) of 0.96 for detecting 23 different modification types in Escherichia coli and Saccharomyces cerevisiae.
• Unsupervised Clustering: Performs unsupervised clustering for tentative classification of PTMs to characterize modification patterns.
• Versatility: Applicable to diverse samples including severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and human transcript data and supports analysis of both native and in vitro unmodified raw RNA reads.

Scientific Applications:

• PTM identification: Detection and mapping of multiple RNA modification types from Nanopore DRS data.
• Comparative analysis: Comparison of native and in vitro unmodified RNA raw reads to identify modification-induced signal deviations.
• Epitranscriptomic profiling across organisms: Characterization of RNA modification landscapes in samples from organisms such as Escherichia coli, Saccharomyces cerevisiae, SARS-CoV-2, and human transcripts.

Methodology:

Analyzes Nanopore electric current signal deviations using a convolutional neural network framed as a Deep One-Class Classification problem to identify PTMs as anomalies.