Establishing a Dorado-Based Nanopore Pipeline for Single-Molecule Modified Base Detection

Jisoo Kim

Mentor: Kazue Hashimoto-Torii, Ph.D. Principal Investigator, Center for Neuroscience Research Children's Research Institute, Children's National Medical Center.

Date/Time: December 9th, 2025 at 12:00 PM.

Abstract: Nanopore sequencing enables direct detection of DNA base modifications from raw ionic current signals, providing single-molecule resolution of epigenetic marks and DNA damage. To prepare for an upcoming experiment aimed at identifying DNA adducts, I developed and validated a complete computational pipeline using publicly available Oxford Nanopore datasets containing 5mC/5hmC methylation as a pilot test system. Raw POD5 files were basecalled on an HPC GPU node using Dorado with a modified-base Remora model, generating aligned BAM files containing MM/ML tags that encode per-base modification probabilities. Downstream processing with modkit produced per-CpG methylation profiles, which were converted into genome-browser–ready tracks and visualized alongside reference annotations. Single-read modification events were further examined in IGV to confirm consistency between MM/ML tags, modkit summary calls, and genome-level methylation patterns.

This validated workflow demonstrates accurate GPU-accelerated modified-base calling, correct interpretation of MM/ML tags, and robust extraction of site-specific modification frequencies. Critically, every component—POD5 handling, Dorado GPU execution, MM/ML parsing, modkit quantification, and IGV/UCSC visualization—is directly transferable to my planned DNA adduct detection experiment, where an adduct-specific Remora model will replace the methylation model. By establishing this computational framework in advance, I ensure readiness for high-confidence detection and interpretation of single-molecule adduct signatures in my upcoming biological samples.