Dynamont: A comprehensive cross-species comparison of ONT segmentation tools

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• Oxford Nanopore Technologies (ONT) sequencing offers direct, long-read DNA and RNA sequencing, preserving nucleotide modifications.
• Accurate basecalling relies on deep neural networks to translate raw nanopore signals into nucleotide sequences.
• Challenges in signal segmentation include signal variability, low-quality reads, and nucleotide modifications, impacting downstream analysis.

Purpose of the Study:

• To conduct a large-scale comparative study of signal segmentation tools for ONT sequencing data.
• To evaluate the performance of five segmentation tools, including the novel tool Dynamont.
• To assess segmentation accuracy across diverse biological data sets and identify optimal tools for specific applications.

Main Methods:

• Applied five segmentation tools to 16 ONT-sequenced data sets from various life forms.
• Segmented a total of 160,000 reads.
• Evaluated tool performance using 12 signal and downstream assembly metrics, aggregated into a comprehensive score.

Main Results:

• No single tool demonstrated superior performance across all tested data sets.
• Dynamont consistently outperformed other tools on nearly all RNA data sets based on aggregated metric scores.
• Performance on DNA data sets was more variable, with mixed results observed among the evaluated tools.

Conclusions:

• Selecting the appropriate segmentation tool requires careful consideration and normalization of evaluation metrics.
• The choice of segmentation tool is a critical step for accurate ONT signal analysis.
• Dynamont presents a promising option for RNA data segmentation, while DNA data segmentation may require tool-specific optimization.