A reproducible analysis pipeline for pH measurements derived from fluorescence microscopy of A. thaliana root tissue.

qbic-pipelines/root-tissue-analysis is a bioinformatics best-practice pipeline to analyze pH measurements from root tissue samples of A. thaliana., these measurments are derived from fluorescence microscopy images. This pipeline aims to analyze pH measurments to validate the acid-growth hypothesis, which explains the expansion of cells in root tissue. This acid-growth pathway model needs substantial pH measurement data for validation, however this type of data generation is time consuming, since manual annotation of ROIs is a mayor bottle-neck. To mitigate this issue, the pipeline provides automatic, multi-class tissue segmentation (5 clases) using U-Net models, previously trained on a dataset generated and annotated by experienced plant biologists (https://github.com/qbic-pipelines/root-tissue-segmentation-core/).

This pipeline was created using nf-core tools and aims to adhere to its reproducibility standards. The pipeline is built using Nextflow, a workflow tool to run tasks across multiple compute infrastructures in a very portable manner. It uses Docker/Singularity containers making installation trivial and results highly reproducible.

• Fluorescence microscopy image files, in .czi or .ome.tif (OME-TIFF) file format. Images of 512x512 pixels in size, and acquired with a target pixel size of 0.415133 µm, each with 4 channels in the following order:

1. Fluorescence signal obtained by excitation at 405 nm
2. Brightfield image for excitation at 405 nm
3. Fluorescence signal obtained by excitation at 458 nm
4. Brightfield image for excitation at 458 nm

• Metadata table in .csv format, with 3 columns (filename, treatment, breeding line), e.g.:

Filename,Treatment,Breeding Line
Image 70,Mock,bri1-301
Image 71,Mock,bri1-301
Image 114,BL,bri1-301
Image 115,BL,bri1-301

Sample input data: Testdata

• Brightfield and ratiomeric images with segmentation masks. Both in .tiff and .npy formats, and integrated as channels within OME-TIFF image files (.ome.tif format)
• Uncertainty and interpretability maps in .ome.tif format
• Average ratio table in .tsv format
• Pipeline report in HTML format

1. Fiji macro for ratiomeric image conversion ([RATIOCONV])

   • RATIOCONV container: https://hub.docker.com/r/qbicpipelines/rtaratioconv

2. Root tissue segmentation. ([ROOTSEG]). This prediction module implements the Monte Carlo Dropout procedure (https://arxiv.org/abs/1506.02142) to calculate prediction uncertainty (uncertainty maps). The number of Monte Carlo samples is set by default to T=10. Additionally, this module uses the Guided Grad-CAM algorithm (https://arxiv.org/abs/1610.02391) to compute visualizations of input feature importance (interpretability maps), as implemented by the Captum library (https://captum.ai/).

   • Segmentation prediction module (mlf-core): https://github.com/qbic-pipelines/rts-prediction-package/
   • ROOTSEG container: https://hub.docker.com/r/qbicpipelines/rtarootseg

3. Export output images in OME-TIFF format ([OMEOUT])

4. Calculate statistics and write pipeline report ([rtastat])

   • rtastat container: https://hub.docker.com/r/qbicpipelines/rtastat

1. Install Nextflow (>=22.10.1)

2. Install any of Docker, Singularity, Podman, Shifter or Charliecloud for full pipeline reproducibility (We strongly recommend using Docker, please only use Conda as a last resort; see docs)

3. Download the pipeline and test with profiles:

   • To test with a local dataset in .czi format, stored in root-tissue-analysis/assets/testdata.tar.gz:

   nextflow run . -profile test_local,docker

   • To test with a local dataset in OME-TIFF format (.ome.tiff), stored in root-tissue-analysis/assets/testdata_ome.tar.gz:

   nextflow run . -profile test_local_ome,docker

   • To test with a remote dataset in .czi format (stored at https://zenodo.org/record/8639352):

   nextflow run . -profile test_remote,docker

4. Download the pipeline and test it on a minimal dataset with a single command:

   nextflow run qbic-pipelines/root-tissue-analysis -profile test,<docker/singularity/podman/shifter/charliecloud/conda/institute>

   • Please check nf-core/configs to see if a custom config file to run nf-core pipelines already exists for your Institute. If so, you can simply use -profile <institute> in your command. This will enable either docker or singularity and set the appropriate execution settings for your local compute environment.
   • If you are using singularity then the pipeline will auto-detect this and attempt to download the Singularity images directly as opposed to performing a conversion from Docker images. If you are persistently observing issues downloading Singularity images directly due to timeout or network issues then please use the --singularity_pull_docker_container parameter to pull and convert the Docker image instead. Alternatively, it is highly recommended to use the nf-core download command to pre-download all of the required containers before running the pipeline and to set the NXF_SINGULARITY_CACHEDIR or singularity.cacheDir Nextflow options to be able to store and re-use the images from a central location for future pipeline runs.
   • If you are using conda, it is highly recommended to use the NXF_CONDA_CACHEDIR or conda.cacheDir settings to store the environments in a central location for future pipeline runs.

5. Start running your own analysis!

   nextflow run qbic-pipelines/root-tissue-analysis -profile <docker/singularity/podman/shifter/charliecloud/conda/institute> --input folder

The qbic-pipelines/root-tissue-analysis pipeline comes with documentation about the pipeline usage, parameters and output.

qbic-pipelines/root-tissue-analysis was originally written by Julian Wanner and Luis Kuhn Cuellar.

If you would like to contribute to this pipeline, please see the contributing guidelines.

An extensive list of references for the tools used by the pipeline can be found in the CITATIONS.md file.

Ewels PA, Peltzer A, Fillinger S, Patel H, Alneberg J, Wilm A, Garcia MU, Di Tommaso P, Nahnsen S. The nf-core framework for community-curated bioinformatics pipelines. Nat Biotechnol. 2020 Mar;38(3):276-278. doi: 10.1038/s41587-020-0439-x. PubMed PMID: 32055031.

Heumos, L., Ehmele, P., Cuellar, L. K., Menden, K., Miller, E., Lemke, S., Gabernet, G., & Nahnsen, S. (2023). mlf-core: a framework for deterministic machine learning. Bioinformatics, 39(4). doi: 10.1093/bioinformatics/btad164.

Barbez, E., Dünser, K., Gaidora, A., Lendl, T., & Busch, W. (2017). Auxin steers root cell expansion via apoplastic pH regulation in Arabidopsis thaliana. Proceedings of the National Academy of Sciences of the United States of America, 114(24), E4884–E4893.