In this module, we create LoadData CSVs with paths to IC functions per channel to use in CellProfiler and segment and extract features from single-cell compartments. We also include thresholds for determining good versus poor quality images as calculated in the previous module to avoid processing poor quality images.
It took approximately 5 days to segment and extract morphology features from single-cell compartments for all six Round_1_data pilot plates, where there were approximately 1,200-2,000 image sets per plate to process. We are using a Linux-based machine running Pop_OS! LTS 22.04 with an AMD Ryzen 7 3700X 8-Core Processor with 16 CPUs and 125 GB of MEM.
To process all six plates from Round_2_data pilot plates, the amount of time it took to run each plate varied by size.
The smallest plate took approximately 12 hours to run and the largest look 2 days and 6 hours to run.
Each plate is ran in parallel but as independent sbatch processes.
We see a significant decrease in computational time when processing as whole plates on HPC.
It only takes about 30 seconds to run generate LoadData CSVs with illum paths.
To generate LoadData CSVs and run the CellProfiler segmentation and feature extraction pipeline, use the command below:
# Make sure you are in the 2.feature_extraction directory
source cp_analysis_local.shWhen running on an HPC, you will only need to call the parent script.
The parent script will call sbatch processes per plate.
sbatch cp_analysis_local.sh