Modena is a nanopore-based computational method for detecting a wide spectrum of epigenetic and epitranscriptomic modifications.

It uses an unsupervised learning approach, namely resampling of nanopore signals followed by the Kuiper test. Unlike other unsupervised tools, classification is performed by 1D clustering of scores into two groups.

To install and use Modena, you need at least Python 3.10 and the Poetry package manager. Then run the following commands:

$ git clone https://github.com/sbidin/modena.git
$ cd modena
$ poetry install
$ poetry run python -m modena --help # See options.
$ poetry --directory path/to/modena/dir/ run python -m modena # Run outside modena dir.

Both datasets need to be supplied in blow5 or slow5 format, alongside their f5c resquiggle output tsv files. If your dataset is in single/multi fast5 format, or pod5 format, you can apply conversions using one of the following tools:

• single-fast5 to multi-fast5: ont_fast5_api
• multi-fast5 to blow5/slow5: slow5tools
• pod5 to blow5/slow5: blue-crab

To resquiggle your data with f5c, install f5c and run the resquiggle command:

$ f5c resquiggle data.fastq data.blow5 > resquiggled.tsv

Both datasets (in this case a and b) need a blow5 or slow5 file and a corresponding f5c-resquiggled tsv file.

$ poetry run python -m modena -a a.blow5 -ax a.tsv -b b.blow5 -bx b.tsv -o out.tsv
$ poetry run python -m modena --help # See here for more options.

Modena outputs a simple tsv file with four columns:

• position, int, 1-based
• coverage, int, a count of all reads that contributed to the signal
• distance, float, a two-sample Kuiper-test-based measure (a distance sum)
• label, str, "pos" or "neg", separating positions into two clusters