COMET

Cost-function Optimized Maximal overlap drift EsTimation

Introduction

COMET is a software package designed to correct drift in single molecule localization (SMLM) datasets with a high spatial and temporal resolution.

Usage

COMET Website

The easiest way to use COMET is to go to our dedicated website, upload your file and let it run on our machine.

Step 1

Bring your dataset in the Thunderstorm dataformat:

• .csv file, comma separated
• the following headers: "frame", "x [nm]", "y [nm]" ("z [nm]" for 3D datasets)
  • -> the "" are part of the header!
  • the file can have more columns/headers, but the above-mentioned have to be there
• check the Thunderstorm page for reference

Step 2

Go to https://www.smlm.tools, press upload and select your correctly formatted .csv file

Step 3

After the upload is finished, you have to specify the dimension of the dataset, the segmentation method and the segmentation parameter depending on the segmentation method (s. Segmentation Methods for more detailed explaination). Press run.

• Tip #1: If you're unsure about the segmentation, check the "keep file for subsequent analysis" checkbox, to be able to run the analysis again with different parameters, without the need to upload the file again

• Tip #2: After the successful upload of the file, the server will check how many jobs are currently queued, it's that's alot, check out the other methods to use the COMET analysis described later.

Step 4

After your job is done, a diagram showing the drift curve should appear, together with a download button. Press the download button to automatically download a .csv file containing the drift estimates.

Google Colab Notebook

Click on the google colab badge at the top of this README and follow the instructions of the notebook

Docker

Build

docker build -t comet ./Python_interface/

Run

docker run --gpus all -v <data_path>:/data comet python -m docker_entrypoint --localizations_path <loc_path> --param_path <param_path> --output_path <output_path> where:

• <data_path> is the folder where input/output data is found
• <loc_path> is the path of the .csv localizations file
• <param_path> is the path of a .yaml file with the algorithm parameters (see Google Colab):
  • "dataset_dimension" (2 or 3)
  • "segmentation_method" (one of "time", "locs", or "frames")
  • "segmentation_parameter"
  • "max_drift_nm"
  • "initial_gaussian_scale_nm"
• <output_path> is the path where the output files go to
• <loc_path>, <param_path>, and <output_path> are relative to /data (/data/...)

Additional Information

Segmentation Methods

3 options to segment the dataset:

1. segment by number of time windows:

this divides the dataset in equal parts containing a fixed number of localizations, you can specify the number of parts.

2. segments by number of locs per window

similar to option 1) but here you have to specify the number of localizations per window and the resulting number of segments is calculated.

3. segment by number of frames per window:

this will use the information provided by the dataset and split the dataset in unequal parts, where each part contains all the localizations witin a number of frames that you specify.