This repository contains source code, scripts, and instructions to reproduce the results presented in FLAG-X: Hybrid machine learning workflows for automated gating of clinical flow cytometry data (doi).
For a standalone implementation of FLAG-X, please refer to https://github.com/bionetslab/FLAG-X.
FLAG-X version 0.2.1 was used to generate the results presented here.
All environments used for the analysis are in ./envs. Details on which environment to use are in the docstring of each script file.
-
The immune status (Imstat) and lymphoma (Lt1, Lt2, Lt1b, Lt2b) datasets are not publicly available due to data protection regulations and ethical restrictions.
-
Flowcyt:
- Download: https://cuicloud.unige.ch/index.php/s/55PHBLEynrp5pN8
- Data processing instructions: https://github.com/VIPER-GENEVA/FlowCyt-Classification-Benchmark.
- Publication: Bini et al., 2024, https://raw.githubusercontent.com/mlresearch/v248/main/assets/bini24a/bini24a.pdf
-
BLAST110:
- Download: https://zenodo.org/records/11046402
- Publication: Mocking et al., 2024, https://doi.org/10.1038/s43856-024-00700-x
# Todo
Generated results are saved to ./results. Results data necessary to reproduce the figures is provided in this repo, raw outputs such as saved models are not provided due to the memory constraints of GitHub.
Generated figures are saved to ./figures.
The ./validation package contains all modules that provide functionality for validation:
gatingprovides Scikit-learn-compatible wrappers for established flow cytometry classification methods:gatemeclass.pyimplementsGateMeClassClassifier, which wraps the GateMeClass method (Caligola et al., 2024, https://doi.org/10.1093/bioinformatics/btae322).dgcytof.pyimplementsDGCyTOFClassifier, which wrap the DGCyTOF method (Cheng et al., 2022, https://doi.org/10.1371/journal.pcbi.1008885.
Claude Opus 5.5 (Anthropic) was used as a coding assistant during the development of this repository. The authors reviewed and validated the resulting code.