sqY2H Screen Quantification

Image-analysis pipeline for converting composite plate images into per-well measurements of colony growth. The workflow corrects positional and geometric variation between images, fits well grids, segments colonies, assigns them to wells, and exports continuous intensity-based measurements. RGB-derived measurements can also support inspection for potential contamination.

Workflow

Run the numbered stages in order:

1. Registration (Fiji)

   • composite_plate_mask_generator.ijm: Interactively generate empirical and ideal template masks for the composite-plate layout in Fiji (run once for a given plate layout and acquisition magnification setting).
   • composite_plate_registration_unwarping_cropping.ijm: Rigidly register each image to the empirical template, apply B-spline deformation toward the ideal geometry, crop the plate region, and export normalized images in batch.

2. Grid fitting (Python)

   • grid_fitting.py: Fit an initial well grid by searching over image translation and rotation, then refine individual grid-line positions to account for residual spatial distortions. Gridded and ungridded transformed images are exported for downstream well and colony segmentation.

3. Colony quantification (CellProfiler)

   • colony_segmentation_and_quantification_pipeline.cppipe: Use CellProfiler to enhance and segment colonies, remove unsuitable objects, segment wells from the fitted grid, assign colonies to wells, and export per-well intensity measurements and quality-control overlays.

4. Postprocessing (Python/Jupyter)

   • postprocessing.ipynb: Parse plate and well identifiers, restructure CellProfiler measurements, and calculate RGB fractions and standardized intensity scores.

5. Validation (Python/Jupyter)

   • GT_comparison.ipynb: Compare continuous image-derived measurements with ordinal manual scores and generate validation and RGB-composition plots.

Example data

examples/ contains three anonymized composite 384-well-plate images together with representative intermediate outputs, measurements, and validation results.

The repository also includes validation plots from a larger evaluation set of 138 composite images, each containing up to six 384-well plates. The underlying full image and scoring datasets are not included.

Plate formats

The supplied masks, grid parameters, and CellProfiler settings are configured for 384-well plates. Other layouts, including 96-well plates (of which three input examples are provided), should be adaptable by:

• generating empirical and ideal masks for the new plate geometry;
• changing the grid dimensions and retuning the grid-fitting parameters; and
• adjusting scale-dependent CellProfiler settings, including expected object sizes and the feature size used by EnhanceOrSuppressFeatures.

Requirements

The Python environment used by the grid-fitting, postprocessing, and validation stages can be created with:

conda env create -f python_environment/environment.yml -n sqy2h
conda activate sqy2h

Related work

This pipeline accompanies the preprint below:

Tejura, M., Chen, Y., McEwen, A. E., et al. (2026). A scalable approach to resolving variants of uncertain significance. bioRxiv.