Benchmarking the Nanopore Direct RNA Sequencing based RNA Modification detection tools (NaRMBench)

Dependency

1. Preprocessing

• This script provides a step-by-step pipeline for preprocessing Oxford Nanopore Direct RNA Sequencing data, from basecalling to alignment and signal re-squiggling using Tombo.

2. Model Retraining

• This directory contains retraining pipelines and modified code for six retrainable-in-practise frameworks (m6Anet, SingleMod, EpiNano, TandemMod, Dinopore, NanoSPA).
• To facilitate model retraining, we have modified the original source code of some tools where retraining functionality was not readily available. The modified versions are provided as ZIP files (e.g., 2.retraining/2.1.retrain_m6Anet/modified_code/m6Anet_modified_code.zip), while EpiNano and TandemMod remain unmodified as they already support retraining out-of-the-box.
• Step-by-step tutorials: a detailed retraining tutorial is provided for each framework (e.g., retraining m6Anet) that guide users through the complete retraining process.
• All retrained models are saved in 2.retraining/2.X.retrain_${tool_name}/retrained_models/ for direct use.

3. Model Evaluation

• We provide a standardized pipeline to evaluate the performance of both original and retrained models which can be found in: evaluation pipeline.

4. Downstream Analysis

• This section contains scripts for downstream analyses and figure generation used in the paper. More scripts for visualization (e.g., distribution of predicted sites, replicate consistency, bias assessment, radar plots, and feature comparisons) are available in the 4.downstream_analysis/ directory.
• Refer to the subdirectory README for detailed descriptions, input requirements, and usage examples.

Authors

• Tingting Luo
• Moping Xu
• Miao Wang
• Faying Chen
• Jiejun Shi