Not yet tested
Other Ubuntu releases, RTX models, RAM capacities, driver versions, storage layouts, and custom references require separate prospective testing. Hardware detection alone is not enough to enable an RNA-seq run.
Public alpha installation
Prepare the analysis data and validated environment profile
Before analysis, the user or system administrator must prepare the required containers, reference data, indices, and environment profile. Vendor terms, registry sign-in, and downloads are not handled automatically; review and complete these steps independently.
Tested setup
| Operating system | Ubuntu 24.04.4 LTS running directly on Linux, not through WSL |
|---|---|
| GPU | NVIDIA GeForce RTX 3090 with 24 GiB VRAM and a compatible driver |
| Memory | 128 GB of physical RAM |
| Active storage | A local, high-speed ext4 SSD with sufficient free capacity. Do not use a slow hard drive or network storage as the Nextflow workDir. |
| Container runtime | Docker Engine and NVIDIA Container Toolkit |
| Workflow engine | Java 17 or later and Nextflow 25.04.3 |
| Application | Python 3.11 or later and Harako-GPU v0.1.0-alpha.1 |
| Alignment image | Parabricks 4.6.0-1 with a verified source, version, and runtime image ID |
| Reference data | The fixed GRCh38.p14 / GENCODE 49 reference pack, STAR 2.7.2a index, and the selected Salmon 2.5.1 and/or 1.10.3 index |
| Environment profile | A validated environment profile created by the system administrator, with matching analysis reports and the plugin and container records required by the selected backend |
Download the release
The v0.1.0-alpha.1 wheel and source distribution are available from GitHub Releases.
They contain the Harako-GPU application only. Parabricks, container images, reference sequences, annotations, and STAR/Salmon indices must be provisioned separately.
Install from a fresh source clone
You can also install the application directly from source:
git clone https://github.com/do-shima/Harako-GPU.git
cd Harako-GPU
python3 -m venv .venv
. .venv/bin/activate
python -m pip install --upgrade pip
python -m pip install -e .
harako-gpu --version
harako-gpu doctor --json
harako-gpu ui --no-browserPrepare the required containers and reference data
- Install Docker Engine and NVIDIA Container Toolkit by following their official documentation.
- Obtain Parabricks 4.6.0-1 from official NVIDIA NGC. Review and accept NVIDIA's terms independently before using it.
- Prepare the required task-container images with their pinned versions and checksums. Run preparation checks them and stops before launch if a selected image is missing; it never pulls images automatically.
- Place the fixed reference pack and version-specific indices outside the Git repository.
- Copy the validated environment profile created by the system administrator to
<runtime-root>/host-profiles/ubuntu_native_rtx3090_ram128_v1.json. Creating a plan does not generate or modify this file. - Run
harako-gpu doctor --json, review the reported capabilities, and then create and review a plan without starting an RNA-seq analysis.
This repository does not distribute Parabricks, reference data, indices, plugin binaries, container archives, FASTQ, or BAM files. Their licenses and terms apply independently.
Other environments
Unavailable in the validated configuration
The validated configuration does not support BAM generation with the fixed 64 GB Windows/WSL host, CPU STAR, macOS, or active work on NTFS, fuseblk, SMB, CIFS, NFS, /mnt/c, /mnt/d, or unsafe paths.
Eligibility depends on validated analysis reports, container source and version records, image and reference/index checksums, the execution environment, and local-filesystem checks. A hardware model name alone is not sufficient.