Table of Contents

• About SeqTagger
• Models available
• Dependencies and versions
• License information
• Citation

About SeqTagger

SeqTagger is a super-fast and accurate demultiplexing algorithm for direct RNA nanopore sequencing datasets. It supports the current sequencing chemistry (SQK-RNA004), and the deprecated chemistry (SQK-RNA002). SeqTagger expects reads in standard nanopore sequencing data format (pod5 or fast5).

Running SeqTagger is as easy as:

docker run --gpus all -u $UID:$GID -v `pwd`:/data lpryszcz/seqtagger run -k models/b04_RNA004 -r -i /data/input_directory_with_reads -o /data/outdir

You can see available demultiplexing models by executing

docker run --gpus all -u $UID:$GID -v `pwd`:/data lpryszcz/seqtagger ls -lah models

For more details see docs/DESCRIPTION.md, for example:

• How does SeqTagger work?
• How many barcodes are supported?
• Running SeqTagger
• Benchmarking of b96_RNA004
• Available models

Models available

Please note that trained models are back-compatible. If you use SeqTagger v2+, all previous models (v1) and latest models (v2) will work.

Chemistry	RNA biotype	Number of barcodes	Model
RNA004	mRNA (or in vitro polyadenylated RNA)	4	b04_RNA004
		13	b13_RNA004_mRNA
		96	b96_RNA004
	tRNA	7	b07_RNA004_tRNA

If you're still interested in demuxing RNA002 runs:

Chemistry	RNA biotype	Number of barcodes	Model
RNA002	mRNA (or in vitro polyadenylated RNA)	4	b04_RNA002
		96	b96_RNA002
	tRNA	4	b04_RNA002_tRNA

Barcode sets

Please note, barcode sets change depending on the model:

** a) b04 models (b04_RNA002, b04_RNA004 and b04_RNA002_tRNA) models are demuxing the same 20nt- barcodes as DeePlexiCon, listed HERE.

Barcode 1: GGCTTCTTCTTGCTCTTAGG Barcode 2: GTGATTCTCGTCTTTCTGCG Barcode 3: GTACTTTTCTCTTTGCGCGG Barcode 4: GGTCTTCGCTCGGTCTTATT

** b) b96 models (b96_RNA002 and b96_RNA004) are using the 37-nt barcodes, listed HERE.

** c) b07 models (b07_RNA004_tRNA) are using a mix of 20nt and 37nt barcodes listed HERE.

Dependencies and versions

You'll need CUDA-compatible (Nvidia) GPU and CUDA v10 or newer installed in your system supporting half-precision (float16). All Nvidia GPUs released from 2019 onward should work without any issues.

Additionally, you'll need to install docker and NVIDIA Container Toolkit.

Versions tested:

Software	Version
CUDA	10, 11, 12
Docker	25+
Nvidia Container Toolkit	1.14

License Information

This project is licensed under the Creative Commons Attribution-NonCommercial 4.0 International License (CC BY-NC-ND 4.0), available here, with the exception of the bonito module, which retains its original license. The full text of the licenses, including modified code, can be found in the bonito directory.

License Dependencies:

• ONT 1.0: bonito
  • Licensed under the Oxford Nanopore Technologies Public License 1.0. Full license text available at ONT 1.0 License.
• MPL 2.0: pod5, ont_fast5_api
  • Licensed under the Mozilla Public License 2.0. Full license text available at MPL 2.0 License.
• BSD 3-Clause: pandas, seaborn, joblib,
  • Licensed under the BSD 3-Clause License. Full license text available at BSD 3-Clause License.
• MIT: mappy, pysam, numpy
  • Licensed under the MIT License. Full license text available at MIT License.
• OTHER: pytorch, numpy
  • Full license text for pytorch is available at pytorch License.
  • Full license text for numpy is available at numpy License.

Please ensure compliance with each license's terms and conditions.

LPP, GD and EMN have filed patent applications (EP24382340 and EP24383144) based on this work at the European Patent Office.

Citation

If you found this work helpful, please cite:

Pryszcz LP*#, Diensthuber G*, Llovera L, Medina R, Delgado-Tejedor A, Cozzuto L, Ponomarenko J and Novoa EM#. Rapid and accurate demultiplexing of direct RNA nanopore sequencing datasets with SeqTagger. Genome Research 2025