A codebase for training and evaluating of predicting phospholipidosis readout from Cell Painting data and molecular descriptors:
- CellProfiler & DeepProfiler morphological feature pipelines
- Compound embedding & classification workflows
- Multimodal fusion of image + chemical data
- Conformal prediction for uncertainty estimation
This project is under development.
General scripts
classification.py
— train single‐cell classifiersgenerate_crops.py
— extract and load image crops for CP pipelinerun_training.nf
— Nextflow wrapper for classification
/CP – CellProfiler (Petter)
classification.py
generate_crops.py
Grit_check.ipynb
run_training.nf
/Compounds – Chemical embeddings & models (Petter + Benjamin)
compound_classification.py
compounds_embedding.ipynb
/DP – DeepProfiler (Petter + Benjamin)
DP_exploration.ipynb
DP_exploration_conformal.ipynb
DP_exploration_conformal_site.ipynb
DP_regression.ipynb
/multimodal – Image + compound fusion (Benjamin)
dataset.py
model.py
,model_v2.py
train.py
optuna_tuning.py
&sweep_config.yaml
wandb_sweep.py
utils.py