Skip to content

Releases: LooseLab/ROBIN

ROBIN model assets v0.5

03 Apr 20:41

Choose a tag to compare

Overview
ROBIN 0.5 is a major release. It brings the Little John orchestration work into the main line, moves the stack to Python 3.12, consolidates the conda layout, and adds substantial GUI, workflow, SNP, MGMT, CNV, fusion, and reporting changes. Treat this as a full reinstall: do not reuse an older ROBIN or little_john conda environment.

Previous tag: v0.1.7 → This release: v0.5

Upgrade requirements (read first)
New conda environment required. Use a single env from robin.yml (env name robin). Older split env files (robin_0_8.yml, robin_py12.yml, robin_osx.yml, legacy Python 3.9 robin.yml) are superseded.
Python 3.12 is the minimum supported version.
Linux: prior optional conda_env_hooks/ / setup_conda_ld_path.sh helpers are removed. If you hit libstdc++-related issues, set LD_LIBRARY_PATH as documented in the README (Common issues) and/or use robin_linux_extras.yml or install libstdcxx-ng / libgcc-ng as needed (plain conda env create may not apply # [linux] selectors in robin.yml automatically).
Parquet / methylation filters: CpG filter handling for .txt filters (e.g. parquet_filter.txt) was corrected (1-based Illumina → 0-based BED). Rebuild existing parquet (remove parquet and re-run bed_conversion) for trustworthy methylation counts after upgrading.
ClinVar bundled resources were updated; refresh assets as per your usual ROBIN update process.
Highlights
Platform & packaging

Version 0.5 across the application and packaging.
Unified robin.yml conda definition; clearer environment story for maintenance and support.
Rich-styled CLI progress where supported (requires an up-to-date install).
Security & GUI access

All GUI routes require login; unauthenticated users go to /login.
Argon2 password hashing; hash stored under ~/.config/robin/gui_password_hash (Windows: %APPDATA%\robin\gui_password_hash). No plaintext or env-only password.
Password set/confirm at terminal on first run (TTY); headless runs use the web login.
robin password set to set or replace the GUI password (Rich prompts when available).
Session invalidation on server restart / password change so old browser sessions are not trusted.
Quit in the menu only when the browser host is localhost / 127.0.0.1 (hidden for remote/LAN access).
Watched folders & workflow

Watched folders: “smart add” — only subfolders that are fully un-analysed are added; mixed trees are handled with warnings instead of silent success.
Manage watched folders as a dedicated /watched_folders page (replaces the old modal); loading spinner while adding.
Path overlap validation: cannot watch paths that overlap the work directory in unsafe ways.
Job-type filtering for large runs; SNP and MNP-Flex global controls in the GUI.
Large BAMs (ROBIN_PROCESS_LARGE_BAMS): optional processing of BAMs >50k reads with per-file batching and clear launch warnings (not recommended alongside live runs).
Ray memory per job type tuned (e.g. higher for fusion and SNP / Clair3); deduplication extended so SNP is not run twice for the same sample.
Analysis

SNP pipeline: Clair3 with snpEff / SnpSift; snp_analysis and igv_bam job types; GUI table + IGV navigation; IGV-ready BAM creation; queueing from the GUI.
MGMT: methylation classification uses per-read probabilities; robin utils mgmt for CpG summaries from mgmt_sorted.bam (recursive search, TSV output, improved percent reporting).
Classification: scoring and aggregation improved across Sturgeon / NanoDX / PanNanoDX / Random Forest.
CNV (GUI): bin-width selector; chromosome set aligned with reports (chr0–22, X, Y; excludes chrM and odd contigs).
Fusion (GUI): faster refresh when the fusion section is built; optional timing logs for diagnostics.
MNP-Flex: optional Epignostix upload for bedMethyl (MNPFLEX_USER / MNPFLEX_PASS); panel hides empty fields until data exists.
Target analysis always emits targets_exceeding_threshold.bed for downstream SNP use.
Reports & PDFs

Design refresh and layout updates.
Classification plots: seaborn-style top 3 classes; tables in a 2-column classifier layout with compact styling.
CNV / coverage tables and plots refined (layout, fonts, overlap fixes, coverage outlier table).
Fixes

GUI locking and timer/refresh reliability.
Multi-BAM samples: no longer skip valid BAMs due to overly aggressive “already analysed” logic per BAM.
Dependencies (not exhaustive)

argon2-cffi, seaborn, Rich (as above).
ClinVar VCF + index updated.
Documentation & support
Internal notes added for fusion GUI timing and methylation extraction comparisons (e.g. FUSION_GUI_DELAY_ANALYSIS.md, METHYLATION_EXTRACTION_COMPARISON.md).
scripts/compare_mgmt_parquet_methylation.py for validating MGMT vs parquet methylation counts.
Full detail
For a line-by-line list of changes, see CHANGELOG.md in the repository. For installation and troubleshooting, see README.md and the MkDocs site (if published).

The make it stable release.

22 Jan 10:06
d2f9b3c

Choose a tag to compare

Since the previous release we have tidied up the report process, fixed a number of bugs around reading files aligned to hg38 including alts and made a few additional fixes.

From this point forwards we will be maintaining the main branch and all releases will have been tested on a selection of samples.

We have also established a dev branch which will contain new features currently under testing. Unless developing

As with all releases this code is for research use only.

The next release will focus on performance optimisations.

This code and the data contained within it are intended for research use only and should not be used for direct diagnostic purposes. The methylation-based classification and other analyses provided here may be considered by neuropathologists as supplementary information in the context of comprehensive diagnostic assessment, which should include clinical history, radiological findings, and complete histopathological and molecular evaluation. The final interpretation and diagnosis should always be made by qualified healthcare professionals based on all available information.

The authors of this software note that:
Oxford Nanopore Technologies products are not intended for use for health assessment or to diagnose, treat, mitigate, cure, or prevent any disease or condition.

Using this software in any context, research or otherwise, is at the users own responsibility.

What's Changed

Full Changelog: v0.1.0...v0.1.7

Robin Release

12 Sep 06:47
30b443e

Choose a tag to compare

This is the intial release of our ROBIN tool as described in

ROBIN: A unified nanopore-based sequencing assay integrating real-time, intraoperative methylome classification and next-day comprehensive molecular brain tumour profiling for ultra-rapid tumour diagnostics
Simon Deacon, Inswasti Cahyani, Nadine Holmes, Graeme Fox, Rory Munro, Satrio Wibowo, Thomas Murray, Hannah Mason, Mark Housley, Daniel Martin, Abid Sharif, Areeba Patel, Robert Goldspring, Sebastian Brandner, Felix Sahm, Stuart Smith, Simon Paine, and Matthew Loose
doi:10.1101/2024.09.10.24313398
http://medrxiv.org/cgi/content/abstract/2024.09.10.24313398v1

Software is provided as is.

Please report any issues via github issues. We will also be monitoring the discussions forums for general advice. Documentation above and beyond the README file will be coming shortly.

Pull requests are welcome.

preview release

25 Jul 18:50
a3b019d

Choose a tag to compare

preview release Pre-release
Pre-release

This is a test release for an initial build of ROBIN.

What's Changed

New Contributors

Full Changelog: https://github.com/LooseLab/ROBIN/commits/v0.0.6