Built for the Department of Health Northern Territory Pathology Genomics, and for anyone who wants an easy-to-use tool.
A desktop app for typing bacterial isolates from paired Illumina reads, or your own FASTQ reads, on your own computer. It supports your whole genome sequencing work without the burden of the command line. No server is needed: it all runs locally on your computer. It lets you import workflows from Galaxy Australia, and gives you an easy way to install any bioinformatics tool from GitHub, so you can bring the tools your work needs. You can also build your own workflows from these tools.
Have an issue or a problem? I am here to help. I give support and can make the app work with other tools: open an issue.
- One installer for Windows, Mac and Linux. No Nextflow, Java or Python to set up.
- Every tool in its own container at a fixed version, so results are repeatable.
- Every run in one folder with its results and a record of exactly what ran.
- Add tools from GitHub and import workflows from Galaxy.
- Species identification that fits a small computer: Kun-peng is built in beside Kraken2 and Bracken. It uses the same Kraken2 databases in far less memory, with Bracken built in.
Version 2. Version 1 (the browser app started with
./ntg) has ended; its saved workflows, databases and results are kept.
Download the installer for your computer from Releases and run it.
| Computer | File | Notes |
|---|---|---|
| Windows 10/11 | ..._x64-setup.exe |
Makes a Start menu entry and a desktop shortcut. |
| Mac (Intel or Apple silicon) | ..._universal.dmg |
Drag the app to Applications. |
| Linux | ..._amd64.AppImage or ..._amd64.deb |
- Windows: the installer shows "Windows protected your PC": choose More info, then Run anyway. It is not yet signed by a publisher Windows knows (see Code signing policy).
- Mac: the app is signed and checked by Apple, so it opens without a warning.
The tools run in containers, which need a container engine. The app's Setup screen and its Settings page say what is missing.
- Windows: the Linux side of Windows (WSL) with Ubuntu, and Docker inside it. The app has buttons for both: Install WSL (asks for administrator rights and a restart) and Set up Docker. Docker Desktop is not needed.
- Mac: Docker through Colima (free). Set up Docker in Settings opens a Terminal window that installs it with Homebrew (and Homebrew itself when it is missing) and starts it with all of the Mac's memory but 4 GB (kept for the Mac itself). Docker Desktop and Rancher Desktop work too.
- Linux: Docker Engine or Apptainer. On Ubuntu and Debian, Set up Docker in Settings installs Docker Engine in a terminal window; log out and in again afterwards.
Settings also shows the same steps as lines to copy, for doing it by hand.
Start and Stop in the side panel switch Docker on and off (on a Mac with Colima's Linux machine, on Windows with the Linux side, WSL), so it takes memory only while you use it. On Mac and Linux Docker does not start with the app: press Start. Stop asks first, and is refused while a run is going. Quit in the tray icon's menu does the same as Stop on Windows (the Linux side) and on a Mac (Colima), unless a run is still going.
About 10 GB of disk for the reference databases, plus room for runs.
Memory: the computer needs at least 16 GB, on Windows, Mac and Linux. With 16 GB some tools may still not work, so more than 16 GB is best. The app keeps 4 GB for the computer itself and gives the rest to the tools; on Windows and Mac the amount can be changed under Settings › Memory for the tools (no file to edit). The app warns before a run about steps that need more than the computer has. When the tools have 16 GB or less, Home suggests Kun-peng for species identification: it needs far less memory than Kraken2 (the suggestion can be dismissed).
Open it from the desktop shortcut or the Start menu. On Windows and Mac, closing the window keeps the app in the system tray (the small icons next to the clock), so it opens again at once. To close it fully, right-click its tray icon and choose Quit. Runs that are going carry on either way, and show again when you open the app.
The app looks for a newer version each time it opens and tells you when one is out. Settings › Updates lists what is new in it, so you can decide. Press Update now (in that message, or in Settings): the app downloads the new version for your computer, installs it over the old one and opens again. Your workflows, tools, databases and runs are kept, and a run in progress carries on. Nothing needs uninstalling.
Versions up to 2.0.1 can't do this yet: install 2.0.2 or newer once from the download page (run the installer over the old version), and the app updates itself from then on. Every update is checked against the app's own signature before it is installed.
| Step | Tool | Version |
|---|---|---|
| Read preprocessing | fastp | 0.24.0 |
| Genome assembly | Shovill (SPAdes) | 1.4.2 |
| Species identification | Kraken2 + Bracken | 2.1.3 / 3.0 |
| Serovar | SISTR | 1.1.1 |
| MLST | mlst | 2.22.0 |
| AMR and plasmids | staramr | 0.11.0 |
| emm typing | emmtyper | 0.2.0 |
| Resistance / virulence genes | ABRicate (NCBI, VFDB) | 1.4.0 |
| AMR (Australian reporting) | abriTAMR | 1.3.0 |
| Shigella typing | ShigaPass typing (ShigaPass + Mykrobe) | 1.0.2 |
| C. diphtheriae tox and MLST typing | C. diphtheriae typing | 1.1.0 |
| Completeness / contamination | CheckM | 1.2.3 |
| Assembly QC | QUAST | 5.3.0 |
| Compressed assemblies | Compress contigs (built-in) | 1.0 |
| Summary tables | RDH summary tables (built-in) | Galaxy's output tables, made the same way |
Also built in, ready to use in your own workflows (not part of the RDH workflow):
| Step | Tool | Version |
|---|---|---|
| MLST (PubMLST, with an EnteroBase cross-check) | MLST typing | 0.5.0 |
| S. pyogenes typing: emm type and subtype, emm cluster and pattern, MLST, M1UK lineage, superantigens, resistance genes, FCT region | S. pyogenes typing | 1.0.0 |
| Species identification with little memory, from a Kraken2 database (Bracken built in) | Kun-peng | 0.7.13 |
| Tuberculosis lineage and drug resistance | TBProfiler | 6.7.0 |
| Tuberculosis drug resistance | tbtAMR | 1.0.4 |
Settings match the Galaxy workflow (the original file is in
reference/). Where the app has to differ from Galaxy, it says
so next to the setting:
- Kraken2 database: Standard-8 (2026-06) instead of Galaxy's full Standard (2020) with 2021 Bracken files, so species abundances can differ a little. It needs about 8 GB of memory; switch the Kraken2 step off on a computer that has less. The Databases page also offers Standard-16 (16 GB of memory) and the full Standard database (about 100 GB of memory, for servers): install one, then pick it in the Kraken2 and Bracken steps (Database). In your own workflows, Kun-peng (built in, see Adding tools) can use any of these databases in a few GB of memory. On the Databases page the three are in one Kraken2 box: click it to see each database and its Kun-peng copy.
- CheckM tree: "Auto" uses the reduced tree on computers with less than 40 GB of memory (Galaxy used the full tree).
- New run: pick a workflow and a folder of reads, check the sample pairs, switch steps off for this run if you like, read the warnings for this computer, and start.
- Runs: live progress drawn as the pipeline, one results table per tool, every result file, and what exactly ran. A stopped or failed run carries on from where it was.
- Workflows: the built-in workflow, and your own. The builder lets you add
steps, connect them, change settings and save; every save is a new version.
Import from Galaxy turns a workflow file saved from Galaxy (
.ga) into a workflow here, with the same tools, versions and settings. - Tools: every tool with its version and container. Add a tool brings one in from its GitHub page; Check for newer versions and See other versions add another version of a tool beside the one you have.
- Databases: download the reference databases (a broken download carries on where it stopped).
- Settings: this computer's readiness, where things are kept, updates and how to report a problem.
Reads: a folder of <sample>_R1….gz / <sample>_R2….gz files. The app
pairs the files automatically (_R1/_R2, .R1/.R2, _1/_2 or .1/.2) or
with your own markers, and shows every pair before you start.
Assembled contigs: a workflow can start from FASTA files instead of reads.
In the builder, click the Input box and choose Assembled contigs; a
run then takes a folder of FASTA files (.fasta, .fa, .fna, .fas,
.ffn, also with .gz), one sample per file, named after the file. That is
the place for tools that work on an assembly, such as an MLST tool. Importing
a Galaxy workflow whose input is a list of FASTA files gives such a workflow.
Each run is one folder in Documents › Genomics Workflow Microbials, named with the date and the run's name. Delete the folder to delete the run.
2026-11-10 11-21 My run/
Results/
samples/<sample>/<step>/ each tool's output for each sample
combined/<step>/ tools that run on all samples at once
summary/ Galaxy's output tables
pipeline_info/ provenance.json, tool_versions.yml, trace.tsv
Run details/ the workflow, the plan, the samples and the tools exactly as they ran
README.txt
Your own workflows, the tools you added, extra tool versions and the databases
are in ~/.gwm (on Windows, on the Linux side). Settings shows each
folder and opens it.
Uninstalling removes only the program. To take away the rest first, open Settings › Remove everything, press Show what can be removed and tick what should go. Each item shows how much disk it uses:
- Tool containers: the downloaded containers the tools run in.
- App data: your workflows, added tools, databases and working files.
- Docker and Ubuntu (Windows only): Docker inside the Linux side of Windows, or that whole Ubuntu with everything kept in it.
Nothing is ticked to start with, and you confirm once before anything is removed. It can't be undone. Your runs in Documents are never touched, and nothing is removed while a run is going. Then uninstall the app the usual way.
The app is meant to work with any tool that has a GitHub page.
Some tools beyond the RDH workflow come built in and are ready to use in your own workflows: MLST typing, S. pyogenes typing, Kun-peng, TBProfiler and tbtAMR (tuberculosis drug resistance). Pasting the link of a built-in tool tells you so before anything is added.
Kun-peng identifies species the way Kraken2 does, from the same databases, but holds only a piece of the database in memory at a time, so a database too big for your computer with Kraken2 can still be used. It reads its own copy of a Kraken2 database. The app makes that copy from a Kraken2 database you installed: by itself the first time a Kun-peng step needs it, or ahead of time on the Databases page (Prepare for Kun-peng). The copy takes as much disk as the database. Each works without the other: on the Databases page you can remove the Kraken2 database and keep the Kun-peng copy, or remove the copy. Bracken is built into the Kun-peng step (a switch, on by default). Kun-peng as released classifies more reads than Kraken2, because it lets a read through on a single match; the step's setting Hit groups counted can be set to Kraken2's way for results close to Kraken2's.
MLST typing comes with PubMLST data submitted up to 31 December 2024, which is free. PubMLST gives newer data only to people with their own account: enter your own PubMLST key (a Public data access key or an OAuth key) on the Databases page or on the tool, and the app downloads the current schemes to your computer (about 1.5 hours, once), checks for newer ones when you ask, and lets each workflow step choose which database it types against. The key stays on your computer and is sent to PubMLST only.
S. pyogenes typing types group A Streptococcus assemblies against PubMLST, by PubMLST's own typing rules, into one table with a row per sample: emm type and subtype, emm-like genes, emm cluster and pattern, MLST, M1UK lineage, superantigen and resistance genes and the FCT-region form with its pilin type. Read its Warnings column first: it says what needs checking. Like MLST typing it comes with PubMLST data to 31 December 2024 and can download the current data with your own PubMLST key (a few minutes). It has its own box on the Databases page and keeps its own key: a key entered for MLST typing is not used for it.
Tools › Add a tool, and paste the tool's GitHub link. Then one of two things happens:
- The tool has a tested setup (the list is in
recipes/, TBProfiler so far): it is ready to use at once. - It has none yet: the app reads the tool's name, licence and newest release and opens the editor. Choose a folder of reads and press Set it up for me: the app gets the tool's container, reads the tool's own help, writes the command, tries it on one sample and lists the result files. When that works, the tool is ready. The container is built from the tool's own recipe or environment file when its repository has one (the program versions its authors use); the ready-made one from BioContainers is then the other choice in the editor, and the one used when the repository has neither.
The automatic setup fits tools with the usual options (reads in, a folder or file out), not every tool. When it doesn't fit, the command is yours to change in the editor, which shows the names to use, the tool's help and a try-out on one sample. Or ask for a tested setup: once it is added here, every copy of the app picks it up.
Everything is kept on your computer; nothing is changed on GitHub. See docs/TOOLS.md for the details.
If a tool has no ready-made container, the app builds one on your computer from the tool's repository (several minutes, once; needs Docker). And whatever a tool writes, its results show as tables: a text or JSON report is read into tables with headings, every sample together (this can be switched off per tool on the Tools page).
If a tool will not install or run, or anything else goes wrong:
- Open an issue (needs a free GitHub account), with the tool's GitHub link and what happened, or
- email t.k.business@outlook.com.
Both are also under Settings › Report an issue in the app.
When a run fails, its page has Save a problem report: one text file, saved where you choose, with what is needed to find out why (the app and computer, the workflow, what each step printed and where it stopped). It holds no sequence data, and sample names and the computer's user name are replaced before it is written. Attach it to your issue or email.
When a tool will not set up, the tool's editor has Save a tool report: one text file with how the tool is set up here and what the app said while it built, set up and tried it. It holds no sequence data and no results; the names of the files it was tried on, the computer's user name, email addresses and anything that looks like a key, token or password are replaced.
- Windows: free code signing provided by SignPath.io,
certificate by SignPath Foundation. This has been
applied for. Until it is approved, Windows shows its "unknown publisher"
warning. A release's Windows files may carry the project's own self-signed
certificate (signer
TinusTK, SHA-1 thumbprintF32A5B3A091059542CF1C6196F7A41303F8AB819): where they do, it lets you check a download really is this project's build (right-click the file › Properties › Digital Signatures), but Windows does not know the certificate, so the warning stays. - Mac: signed with the maintainer's Apple Developer ID and notarized by Apple.
- Every installer is built from this repository by GitHub Actions
(
.github/workflows/release.yml), from a tagged commit.
Team roles:
Privacy policy: https://tinustk.github.io/GWM/privacy-policy.html. This program will not transfer any information to other networked systems unless specifically requested by the user or the person installing or operating it, other than the downloads the policy lists (its own updates, tool containers and reference databases), which carry no information about the user or their data.
The app is written in Rust (rust/): gwm-core (tools, workflows, the
compiler), gwm-engine (runs each tool in its container, resume, provenance),
gwm-server (the local API), gwm-cli (the gwm command), rust/gui
(Svelte) and rust/desktop (Tauri). Tools are in tools/, the built-in
workflow in workflows/, the database list in databases/.
cd rust
cargo test --workspace
cargo run -p gwm-cli -- serve --port 8642 --gui gui/dist # after: cd gui && npm ci && npm run buildInstallers are built by .github/workflows/release.yml when a v2* tag is pushed.
If you use this app, please cite this repository (see
CITATION.cff) and the tools it runs (each tool's citation is
shown on the Tools page).
MIT, see LICENSE. The tools the app runs keep their own licences;
the app only pulls their published containers.

