De novo Identification of DNA Modifications Enabled by Genome-Guided Nanopore Signal Processing
Marcus H. Stoiber, Joshua QuickiD, Rob S EganiD, Ji Eun LeeiD, S CelnikeriD, Robert K. NeelyiD, Nicholas James LomaniD, Len A. PennacchioiD, James B. BrowniD
Abstract
Advances in nanopore sequencing technology have enabled investigation of the full catalogue of covalent DNA modifications. We present the first algorithm for the identification of modified nucleotides without the need for prior training data along with the open source software implementation, nanoraw. Nanoraw accurately assigns contiguous raw nanopore signal to genomic positions, enabling novel data visualization, and increasing power and accuracy for the discovery of covalently modified bases in native DNA. Ground truth case studies utilizing synthetically methylated DNA show the capacity to identify three distinct methylation marks, 4mC, 5mC, and 6mA, in seven distinct sequence contexts without any changes to the algorithm. We demonstrate quantitative reproducibility simultaneously identifying 5mC and 6mA in native E. coli across biological replicates processed in different labs. Finally we propose a pipeline for the comprehensive discovery of DNA modifications in any genome without a priori knowledge of their chemical identities.
02 /Specification & Classification
Click any tag to browse matching papers- Citations
- 276
- Organism
- Assay
- Research Focus
- Approach
In-depth summary
Claude Opus 5.5 · xhigh effort · 17 Oct 2026 · from the full text of version 2 (10 Apr 2017) · CC BY-ND 4.0
Base-by-base tests of raw nanopore current, native versus amplified DNA, detect 4mC, 5mC and 6mA without training data, recovering seven methylases' motifs (AUC 0.62–0.97).
Earlier nanopore modification callers work on processed event data and need training sets for each modification and even each sequence context, so they cannot find unknown marks or closely spaced modified and unmodified bases. nanoraw maps raw current to genomic positions and statistically tests native against amplified (unmodified) DNA, so neither training data nor prior knowledge of the chemistry is required.
Key results
ShownWith no training data and no algorithm changes, comparing in vitro methylated with amplified DNA recovered the target motifs of seven methylases making 4mC, 5mC or 6mA, and significance peaked at the methylated base itself.
Amplified E. coli DNA methylated in vitro by M.BamHI (4mC), M.HhaI, M.MpeI, M.SssI (5mC), M.TaqI, M.EcoRI and M.dam (6mA): genome-wide AUC for the known motif 0.62 to 0.97; M.BamHI rises from 0.84 to 0.93 when a discovered degenerate fourth motif position is allowed; for six of seven enzymes, testing the adjacent base instead lowers AUC. · Identification of chemically modified…; Modification Based Sequence Motifs
ShownIn native E. coli K-12 MG1655 sequenced independently in two labs, the strongest de novo motifs were the expected endogenous M.dam (6mA) and M.dcm (5mC) sites.
AUC 0.85 (M.dam) and 0.93 (M.dcm) in Experiment A (98X native/13X amplified strand-specific coverage) vs 0.73 and 0.90 in Experiment B (8X/6X); 97% (A) and 88% (B) of the top 1000 sites lie within 2 base pairs of these motifs. · Endogenous modifications in laboratory strain of…
ShownReproducibility between the two labs' biological replicates is driven mainly by sequencing depth.
The tops of the two rank lists overlap 200 fold over random; overlap of the top 1% of sites rises with coverage beyond 15X and exceeds 32% for sites above 11X; down-sampled Experiment A replicates agree about as well as equal-depth samples from the two labs. · Endogenous modifications in laboratory strain of…
ShownSimple median normalization of the raw current, needing no basecalls or k-mer lookup tables, captures about as much k-mer-driven signal variance as the standard picoamp normalization.
Variance explained by 4-mers: 76% for both picoamp (pA) and median normalization; for 6-mers, 77% (pA) vs 76% (median). Median normalization uses only raw 16-bit data acquisition (DAC) values, centred on the read median, scaled by median absolute deviation (MAD) and clipped at five MAD. · A note on signal normalization; Signal Level Normalization
SuggestedClustering per-site signal-shift signatures separates modification types without knowing their chemistry, a proposed route to discovering unknown marks.
Native-minus-amplified differences within two bases of each significant site, compared by Euclidean distance (shifts up to 2 bases) and projected by multidimensional scaling: the two dominant clusters are said to match known 4mC and 5mC methylases. No cluster-quality metric is reported, and which samples were clustered is unclear. · Modification Clustering; Clustering
Methods, data and reproducibility
nanoraw aligns reads to a genome (BWA-MEM), re-segments raw current to match the reference (genome_resquiggle), median-normalizes it, and tests native against whole-genome-amplified DNA at each base with at least 5 reads per sample using a Mann-Whitney U-test, combining p-values over a window of two bases either side with Fisher's method; top sites feed MEME motif discovery and MDS clustering.
- Ground-truth set: whole-genome-amplified E. coli DNA methylated in vitro by seven methylases (M.BamHI, M.HhaI, M.MpeI, M.SssI, M.TaqI, M.EcoRI, M.dam), plus control samples; nanopore 2D reads · E. coli · n = seven methylated samples + two controls
- Native and amplified DNA, Experiment A (Pennacchio Lab, LBNL, USA) · E. coli K-12 MG1655 · n = 98X native / 13X amplified average strand-specific coverage
- Native and amplified DNA, Experiment B (Loman Lab, Univ. Birmingham, UK), run alongside two other bacterial samples · E. coli K-12 MG1655 · n = 8X native / 6X amplified average strand-specific coverage
- Methods & tools
- nanoraw 0.4.2, BWA-MEM, graphmap, MEME (ZOOPS model), Oxford Nanopore sequencing (2D reads), whole genome amplification, Mann-Whitney U-test, Fisher's method, multidimensional scaling (MDS)
- Code
- https://github.com/marcus1487/nanorawhttps://pypi.python.org/pypi/nanorawhttps://nanoraw.readthedocs.io/
- Software & environment
- nanoraw version 0.4.2 (Python package on PyPI); versions of other software are not stated.
- Reproducibility gaps
- No data availability statement or accession numbers for the nanopore runs; sample preparation, in vitro methylation and the full re-squiggle algorithm are deferred to supplementary methods that are not in this text.
Limitations (6)
- Detection needs the same modification to recur at a position across several independent reads (testing requires at least 5 reads per sample), so single-molecule modification calls are not yet possible. · stated by the authors
- Only template reads were analysed, because complement-read signal is shifted, likely by re-annealing of the DNA helix. · stated by the authors
- Accuracy varies widely across methylases (AUC 0.62 to 0.97), which the authors attribute partly to star activity at near-cognate sites and possibly low in vitro methylase activity. · stated by the authors
- Sequencing depth drives reproducibility: power keeps rising beyond 15X coverage, and the low-coverage Experiment B gave lower AUCs than Experiment A. · stated by the authors
- Endogenous validation uses one well-characterized E. coli strain whose marks were already known; discovery of unknown modifications, eukaryotic genomes, newer nanopore chemistries and RNA are proposed but not tested here. · noted in this summary
- The clustering result is qualitative, and its dominant clusters are named 4mC and 5mC although the endogenous E. coli marks reported are 6mA and 5mC, so it is unclear which samples were clustered. · noted in this summary
Who should read:Read this if you want to detect DNA methylation or other base modifications from raw nanopore signal without training data, or need a tool to map and visualize raw current along a reference genome.
Reproducibility ledger
Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.
Sign in to file a reproducibility report.
Related papers
- NanoMod: a computational tool to detect DNA modifications using Nanopore long-read sequencing data
Qian Liu, Daniela Georgieva, … Kai Wang·BMC GenomicsPeer-reviewed·Feb 1, 2019
- Write and Read: Harnessing Synthetic DNA Modifications for Nanopore Sequencing
Uri Bertocchi, Assaf Grunwald, … Yuval Ebenstein·ACS NanoPeer-reviewed·Nov 3, 2025
- Uncalled4 improves nanopore DNA and RNA modification detection via fast and accurate signal alignment
Sam Kovaka, Paul W. Hook, … Michael C. Schatz·Nature MethodsPeer-reviewed·Mar 28, 2025
- DNA methylation calling tools for Oxford Nanopore sequencing: a survey and human epigenome-wide evaluation
Yang Liu, Wojciech Rosikiewicz, … Sheng Li·Genome biologyPeer-reviewed·Oct 18, 2021
- Methylartist: Tools for Visualising Modified Bases from Nanopore Sequence Data
Seth W. Cheetham, Michaela Kindlova, Adam D. Ewing·bioRxiv·Jul 22, 2021
- Cytosine Variant Calling with High-throughput Nanopore Sequencing
Arthur C Rand, Miten Jain, … Benedict Paten·Nature MethodsPeer-reviewed·Feb 20, 2017