DEGA-106-Segmentation-Metrics

[cornhundred]

Adding a qc module that will contain a iST segmentation metrics method.

[huanlity]

Hi @jaspreetishar, great work! I had some minor comments that is focusing on readability, simplicity, and reproducibility of our codebase. Other than that, can share the data dir data/segmentation_metrics_data so I can do a quick test of the QC notebook? Thanks!

[huanlity]

@jaspreetishar These are minor but can enhance readability, simplicity, and mostly important, reproducibility.

For the string that needs to be repeated multiple times (say more than 2-3 times),it is better to make it a variable so you only need to change one place later if needed. 2. Additionally, use for loop if possible.

For example:
cell_polygon_metadata_files = [f"data/segmentation_metrics_data/inputs/{directories[0]}/cell_metadata.parquet", f"data/segmentation_metrics_data/inputs/{directories[1]}/cell_metadata.parquet", f"data/segmentation_metrics_data/inputs/{directories[2]}/cell_metadata.parquet", None]

and

paths_output_cell_metrics = [f"data/segmentation_metrics_data/outputs/cell_specific_metrics_{dataset_name}-{segmentation_approaches[0]}.csv", f"data/segmentation_metrics_data/outputs/cell_specific_metrics_{dataset_name}-{segmentation_approaches[1]}.csv", f"data/segmentation_metrics_data/outputs/cell_specific_metrics_{dataset_name}-{segmentation_approaches[2]}.csv", f"data/segmentation_metrics_data/outputs/cell_specific_metrics_{dataset_name}-{segmentation_approaches[3]}.csv"]

can be simplified as:
cell_polygon_metadata_files = [ f'{seg_qc_dir}/inputs/{d}/cell_metadata.parquet' for d in directories[:3] ] + [None]

paths_output_cell_metrics = cell_polygon_metadata_files = [ f'{seg_qc_dir}/outputs/cell_specific_metrics_{dataset_name}-{sa}.csv' for sa in segmentation_approaches ]

[huanlity]

Thanks for providing the example data. I am getting an error during the 4th iteration of the for loop, presumably, it is from skin_xenium_default. Do you have any idea?

[jaspreetishar]

The error was due to a faulty transformation matrix. I have fixed it and will share it shortly.

[huanlity]

Thanks for making the effort to simplify the code. In lines with reproducibility, the variable directories can be calculated as following so you don't need to spell them out and next person on a different dataset only need to change the value of variable dataset_name instead of 4 strings: directories = [f"{dataset_name.lower()}_{d.lower().replace('-','_')}" for d in segmentation_approaches]. This one is minor you don't have to change the code but it will be good to get familiar with.

[huanlity]

Does this only support Xenium? What if we are QCing a MERSCOPE dataset?

[jaspreetishar]

I have now added support for Merscope as well