SpotMAX benchmark methods

A unified Python interface to run all the methods we benchmarked in the SpotMAX manuscript on the SpotMAX 3D ground-truth dataset.

Instructions:

1. Clone this repository with the command git clone https://github.com/SchmollerLab/SpotMAX_benchmark.git
2. Download the dataset from here
3. Extract the dataset to the folder SpotMAX_benchmark\data\SpotMAX_gt_dataset
4. Install Cell-ACDC in a dedicated conda environment by following these instructions

To run the analysis, open the terminal, activate the conda environment, navigate to the SpotMAX_benchmark folder you cloned before, and run the following command:

python src\run.py -m <method_name> -c <channel_name>

where you need to replace <method_name> and <channel_name> with one of the following options:

• Methods: Piscis, Big-FISH, Spotiflow, U-FISH
• Channels: mNeon, Cy3, MDN1.

The mNeon channel is the mitochondrial DNA dataset, while Cy3 and MDN1 are the channels of the single-molecule FISH dataset.

The other available flags for the command are the following:

• -s: name of the Spotiflow name to run. Available models are smfish_3d, synth_3d, and 2025-01-26_10-57-34_finetune_from_synth_3d
• -f: force overwriting existing files for the selected method
• -i: inspect results on each image

Once the analysis is completed, each Images folder in the dataset will have the CSV file of the selected method containing the coordinates of the detected spots.

For example, for Piscis, you will have the following file:

"SpotMAX_benchmark\data\SpotMAX_gt_dataset\Yeast_smFISH\MDN1_gene\Position_1\Images\MMY116_2c_ACT1_MDN1_7_s10_MDN1_piscis_coords.csv"