Link Search Menu Expand Document

Directory structure

Upon the full successful completion of a pipeline run, the directory structure will follow Fig. 1A in the mcmicro manuscript:

SchematicDirectory Structure
├── markers.csv
├── raw/
├── illumination/
├── registration/
├── dearray/
├── probability-maps/
├── segmentation/
├── quantification/
└── qc/

The name of the parent directory (e.g., exemplar-002) is assumed by the pipeline to be the sample name.

Table of contents
  1. Directory structure
  2. Raw data
  3. (Optional) Illumination correction
  4. Stitching and registration
  5. (Optional) TMA dearray
  6. Segmentation
  7. Quantification
  8. Quality control

Raw data

At the very minimum, the pipeline expects markers.csv, containing metadata about markers, in the parent directory and raw images in the raw/ subdirectory. The file markers.csv must be in a comma-delimited format and contain a column titled marker_name that defines marker names of every channel:


All other columns are optional but can be used to specify additional metadata (e.g., known mapping to cell types) to be used by individual modules.

The exemplar raw/ files are in the open standard OME-TIFF format, but in practice your input files will be in whatever format your microscope produces. The pipeline supports all Bio-Formats-compatible image formats.

(Optional) Illumination correction

├── markers.csv
├── raw/
└── illumination/
    ├── exemplar-001-cycle-01-dfp.tif
    ├── exemplar-001-cycle-01-ffp.tif
    ├── exemplar-001-cycle-02-dfp.tif
    ├── exemplar-001-cycle-02-ffp.tif
    ├── exemplar-001-cycle-03-dfp.tif
    └── exemplar-001-cycle-03-ffp.tif

Precomputed flat-field and dark-field illumination profiles must be places in the illumination/ directory. If no precomputed profiles are available, mcmicro can compute these using BaSiC. This step is not executed by default, because proper illumination correction requires careful curation and visual inspection of the profiles produced by computational tools. After familiarizing yourself with the general concepts [ 1, 2 ], the profiles can be computed by specifying --start-at illumination during pipeline execution.

Stitching and registration

The first step of the pipeline will aggregate individual image tiles in raw/ along with the corresponding illumination profiles to produce a stitched and registered image, which will be published to the registration/ subdirectory:

├── markers.csv
├── raw/
├── illumination/
└── registration/
    └── exemplar-001.ome.tif

The output filename will be generated based on the name of the project directory.

(Optional) TMA dearray

When working with Tissue Microarrays (TMA), the registration/ folder will contain an image of the entire TMA. Use the --tma flag during pipeline execution to have mcmicro identify and isolate individual core. Each core will be written out into a standalone file in the dearray/ subdirectory along with its mask specifying where in the original image the core appeared:

├── ...
├── registration/
│   └── exemplar-002.ome.tiff
└── dearray/
    ├── 1.tif
    ├── 2.tif
    ├── 3.tif
    ├── 4.tif
    └── masks/
        ├── 1_mask.tif
        ├── 2_mask.tif
        ├── 3_mask.tif
        └── 4_mask.tif

All cores will then be processed in parallel by all subsequent steps. In MCMICRO, TMA dearraying is performed by Coreograph.


Cell segmentation is carried out in two steps. First, the pipeline generates probability maps that annotate each pixel with the probability that it belongs to a given subcellular component (nucleus, cytoplasm, cell boundary). The second step applies standard watershed segmentation to produce the final cell/nucleus/cytoplasm/etc. masks. The two steps will appear in probability-maps/ and segmentation directories, respectively. When there are multiple modules for a given pipeline step, their results will be subdivided into additional subdirectories:

├── ...
├── probability-maps/
│   ├── ilastik/
│   │   └── exemplar-001_Probabilities.tif
│   └── unmicst/
│       └── exemplar-001_Probabilities_0.tif
└── segmentation/
    ├── ilastik-exemplar-001/
    │   ├── cell.ome.tif
    │   └── nuclei.ome.tif
    └── unmicst-exemplar-001/
        ├── cell.ome.tif
        └── nuclei.ome.tif


The final step combines information in segmentation masks, the original stitched image and markers.csv to produce Spatial Feature Tables that summarize the expression of every marker on a per-cell basis, alongside additional morphological features (cell shape, size, etc.). Spatial Feature Tables will be published to the quantification/ directory:

├── ...
├── segmentation/
└── quantification/
    ├── ilastik-exemplar-001_cell.csv
    └── unmicst-exemplar-001_cell.csv

There is a direct correspondence between the .csv filenames and the filenames of segmentation masks. For example, quantification/unmicst-exemplar-001_cell.csv quantifies segmentation/unmicst-exemplar-001/cell.ome.tif.

Quality control

Additional information during pipeline execution will be written to the qc/ directory, by both individual modules and the pipeline itself.

├── ...
└── qc
    ├── params.yml
    ├── provenance/
    │   ├── probmaps:ilastik (1).log
    │   ├── probmaps:ilastik (1).sh
    │   ├── probmaps:unmicst (1).log
    │   ├── probmaps:unmicst (1).sh
    │   ├── quantification (1).log
    │   ├── quantification (1).sh
    │   └── ...
    ├── coreo/
    ├── s3seg/
    └── unmicst/

While the exact content of the qc/ directory will depend on which modules were executed, two sources of information can always be found there:

  1. The file params.yml will contain the full record of module versions and all parameters used to run the pipeline. This allows for full reproducibility of future runs.
  2. The provenance/ subdirectory will contain exact commands (.sh) executed by individual modules, as well the output (.log) of these commands.

The remaining directories will contain QC files specific to individual modules:

  1. When working with TMAs, coreo/ will contain TMA_MAP.tif, a mask showing where in the original TMA image the segmented cores reside.
  2. If UnMicst was used to generate probability maps, unmicst/ will contain thumbnail previews, allowing for a quick assessment of their quality.
  3. After segmentation, two-channel tif files containing DAPI and nuclei/cell/cytoplasm outlines will reside in s3seg/, allowing for a visual inspection of segmentation quality.

After sufficient quality of the outputs has been established and no more parameter tuning is expected, the QC files can be safely deleted. It is recommended to retain params.yml and provenance/ because of their relatively small file size, given that these files enable full reproducibility of a pipeline run.