chore: add DOI and update lock file

README.md

@@ -2,8 +2,10 @@
 
 [![CI](https://github.com/AI4SCR/car-conditional-monge/actions/workflows/ci.yml/badge.svg)](https://github.com/AI4SCR/car-conditional-monge/actions/workflows/ci.yml)
 [![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
+[![DOI](https://zenodo.org/badge/885465953.svg)](https://doi.org/10.5281/zenodo.17526342)
 
-The conditional Monge Gap applied the single cell RNA sequencing data of Chimeric Antigen Receptor T cells. Extension of the [Conditional Monge Gap](https://github.com/AI4SCR/conditional-monge), to include CAR specific dataloaders, embeddings and trainers. Additionally `notebooks` contains Notebooks for generating the figures of the preprint ... and additional analyses. In the `configs` and `scripts` directories are all scripts to replicate the experiments from this preprint.
+
+The conditional Monge Gap applied the single cell RNA sequencing data of Chimeric Antigen Receptor T cells. Extension of the [Conditional Monge Gap](https://github.com/AI4SCR/conditional-monge), to include CAR specific dataloaders, embeddings and trainers. Additionally `notebooks` contains Notebooks for generating the figures of the preprint and additional analyses. In the `configs` and `scripts` directories are all scripts to replicate the experiments from this preprint.
 
 ## Development setup & installation
 We use [poetry](https://python-poetry.org/docs/managing-environments/) as package manager and tested the code in Python 3.10.

carot/utils.py

@@ -182,6 +182,7 @@ def score_transports_and_targets_combinations(all_expr, all_meta):
 
     return scores
 
+
 def monge_get_source_target_transport(
     trainer: "MongeGapTrainer",
     datamodule: "AbstractDataModule",
@@ -268,4 +269,4 @@ def monge_get_source_target_transport(
     all_expr = pd.concat(all_expr).reset_index(drop=True)
     all_meta = pd.concat(all_meta).reset_index(drop=True)
 
-    return all_expr, all_meta
+    return all_expr, all_meta

scripts/chemCPA_MMD_and_WD.py

@@ -1,12 +1,13 @@
-from cmonge.evaluate import log_mean_metrics, log_metrics
-import jax
-import jax.numpy as jnp
-from jaxtyping import PRNGKeyArray
 import pathlib
 import pickle
 from typing import Iterator
+
+import jax
+import jax.numpy as jnp
 import typer
 import yaml
+from cmonge.evaluate import log_mean_metrics, log_metrics
+from jaxtyping import PRNGKeyArray
 
 
 def format_dict(dict_):
@@ -50,7 +51,7 @@ def evaluate_condition(
 def main(config: str, split: str):
 
     # donors = ["D09", "D17"]
-    donors = ["D09"] # in case of dummy data
+    donors = ["D09"]  # in case of dummy data
     metrics = {}
     key = jax.random.key(0)
     with open(config, "r") as f:
@@ -63,9 +64,7 @@ def main(config: str, split: str):
 
     # All drug_dose combis in current OOD results
     # chemCPA condition also contains cell line, but we don't use it
-    current_cars = list(
-        set([c.split("_")[0] for c in predictions_dict.keys()])
-    )
+    current_cars = list(set([c.split("_")[0] for c in predictions_dict.keys()]))
 
     # We gather all cells from same drug-dose combi as in CMonge
     for car in current_cars:
@@ -96,9 +95,7 @@ def main(config: str, split: str):
         target_loader = sampler_iter(all_target, batch_size=512, key=k1)
         transport_loader = sampler_iter(all_transport, batch_size=512, key=k2)
 
-        evaluate_condition(
-            target_loader, transport_loader, degs_idx, metrics[car]
-        )
+        evaluate_condition(target_loader, transport_loader, degs_idx, metrics[car])
         log_mean_metrics(metrics[car])
 
         format_dict(metrics[car])

scripts/chemCPA_ckpt_to_carot_eval.py

@@ -1,26 +1,23 @@
-from chemCPA.data import load_dataset_splits
-
-import yaml
-import pandas as pd
-from pathlib import Path
 import pickle
-import numpy as np
+from pathlib import Path
 
+import numpy as np
+import pandas as pd
 import scanpy as sc
 import torch
 import typer
-from tqdm.auto import tqdm
-
+import yaml
 from chemCPA.data import (
     canonicalize_smiles,
     drug_names_to_once_canon_smiles,
+    load_dataset_splits,
 )
 from chemCPA.embedding import get_chemical_representation
 from chemCPA.model import ComPert
 from chemCPA.paths import CHECKPOINT_DIR
 from chemCPA.train import bool2idx, compute_prediction, repeat_n
-
 from cmonge.metrics import average_r2, compute_scalar_mmd, wasserstein_distance
+from tqdm.auto import tqdm
 
 device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 
@@ -47,7 +44,6 @@ def load_smiles(dataset, key_dict):
     drugs_names_unique_sorted = np.array(sorted(drugs_names_unique))
     canon_smiles_unique_sorted = drugs_names_unique_sorted
 
-
     return canon_smiles_unique_sorted
 
 
@@ -69,11 +65,11 @@ def load_model(config, canon_smiles_unique_sorted, checkpoint_dir=CHECKPOINT_DIR
     if len(dumped_model) == 3:
         print("This model does not contain the covariate embeddings or adversaries.")
         state_dict, init_args, history = dumped_model
-        COV_EMB_AVAILABLE = False
+        cov_emb_available = False
     elif len(dumped_model) == 4:
         print("This model does not contain the covariate embeddings.")
         state_dict, cov_adv_state_dicts, init_args, history = dumped_model
-        COV_EMB_AVAILABLE = False
+        cov_emb_available = False
     elif len(dumped_model) == 5:
         (
             state_dict,
@@ -82,7 +78,7 @@ def load_model(config, canon_smiles_unique_sorted, checkpoint_dir=CHECKPOINT_DIR
             init_args,
             history,
         ) = dumped_model
-        COV_EMB_AVAILABLE = True
+        cov_emb_available = True
         assert len(cov_emb_state_dicts) == 1
     append_layer_width = (
         config["dataset"]["n_vars"]
@@ -99,7 +95,7 @@ def load_model(config, canon_smiles_unique_sorted, checkpoint_dir=CHECKPOINT_DIR
         device=device,
     )
     model = model.eval()
-    if COV_EMB_AVAILABLE:
+    if cov_emb_available:
         for embedding_cov, state_dict_cov in zip(
             model.covariates_embeddings, cov_emb_state_dicts
         ):
@@ -142,14 +138,14 @@ def compute_pred(
     dosages=[1e4],
     cell_lines=None,
     genes_control=None,
-    use_DEGs=True,
+    use_degs=True,
     verbose=True,
 ):
     # dataset.pert_categories contains: 'celltype_perturbation_dose' info
     pert_categories_index = pd.Index(dataset.pert_categories, dtype="category")
 
     cl_dict = {
-        torch.Tensor([1, 0 ]): "D09",
+        torch.Tensor([1, 0]): "D09",
         torch.Tensor([0, 1]): "D17",
     }
 
@@ -250,7 +246,7 @@ def compute_pred(
 
         mean_pred = mean_pred.cpu().numpy()
         y_true = y_true.cpu().numpy()
-        if use_DEGs:
+        if use_degs:
             drug_r2[cell_drug_dose_comb] = average_r2(
                 y_true[:, idx_de], mean_pred[:, idx_de]
             )
@@ -270,9 +266,7 @@ def main(config_path, split):
     cell_lines = ["D09", "D17"]
     dataset, key_dict = load_dataset(config)
     config["dataset"]["n_vars"] = dataset.n_vars
-    canon_smiles_unique_sorted = load_smiles(
-        dataset, key_dict
-    )
+    canon_smiles_unique_sorted = load_smiles(dataset, key_dict)
 
     # Load dataset
     data_params = config["dataset"]["data_params"]
@@ -294,7 +288,7 @@ def main(config_path, split):
         genes_control=datasets["test_control"].genes,
         dosages=dosages,
         cell_lines=cell_lines,
-        use_DEGs=True,
+        use_degs=True,
         verbose=True,
     )
 

scripts/create_chemCPA_model_per_car_configs.py

@@ -19,10 +19,10 @@ def change_and_save_configs(
     cars = [car.rstrip() for car in cars]
 
     for car in cars:
-        config = base_config.copy()        
-        config["dataset"]["data_params"]["dataset_path"]= adata_path
+        config = base_config.copy()
+        config["dataset"]["data_params"]["dataset_path"] = adata_path
         config["dataset"]["data_params"]["split_key"] = f"{car}_ID"
-    
+
         exp_path = f"{exp_base_dir}/{car}/"
         try:
             os.makedirs(exp_path)
@@ -42,15 +42,21 @@ def change_and_save_configs(
 
 
 if __name__ == "__main__":
-    data_name = "CD4" # CD4, CD8
-    base_config_path = Path(f"/Users/alicedriessen/Projects/car-conditional-monge/configs/chemCPA_{data_name}_ID_config_dummy_donor.yml")
+    data_name = "CD4"  # CD4, CD8
+    base_config_path = Path(
+        f"/Users/alicedriessen/Projects/car-conditional-monge/configs/chemCPA_{data_name}_ID_config_dummy_donor.yml"
+    )
     configs_save_path = Path(
         f"/Users/alicedriessen/Projects/car-conditional-monge/configs/chemCPA_per_car/{data_name}/",
     )
 
-    exp_base_dir = Path(f"/Users/alicedriessen/Box/LegacyFromOldColleagues/Alice/CAR_Tcells/Model/chemCPA/model_per_car/{data_name}")
+    exp_base_dir = Path(
+        f"/Users/alicedriessen/Box/LegacyFromOldColleagues/Alice/CAR_Tcells/Model/chemCPA/model_per_car/{data_name}"
+    )
     adata_path = f"/Users/alicedriessen/Box/LegacyFromOldColleagues/Alice/CAR_Tcells/Model/OT/{data_name}_chemCPA_anno.h5ad"
-    car_variants = Path("/Users/alicedriessen/Box/LegacyFromOldColleagues/Alice/CAR_Tcells/Model/OT/CAR_variants.txt")
+    car_variants = Path(
+        "/Users/alicedriessen/Box/LegacyFromOldColleagues/Alice/CAR_Tcells/Model/OT/CAR_variants.txt"
+    )
 
     for d in [configs_save_path, exp_base_dir]:
         try:

scripts/train_chemCPA.py

@@ -1,6 +1,7 @@
 import pickle
-import yaml
+
 import typer
+import yaml
 from chemCPA.experiments_run import ExperimentWrapper
 
 
@@ -13,7 +14,6 @@ def main(config_path):
     ) as f:
         args = yaml.safe_load(f)
 
-
     exp.seed = 1337
     # loads the dataset splits
     exp.init_dataset(**args["dataset"])

scripts/train_ood_car.py

@@ -36,7 +36,9 @@ def train_conditional_monge(config_path: Path):
         data = config.data.file_path.split("/")[-1][:-5]
         embed = config.model.embedding.name
         car = ""
-        logger_path = Path(f"/Users/alicedriessen/Box/CAR_Tcells/Model/conditional-monge/experiments/cmonge_ood/41BB_OOD/{data}/logs.yml")
+        logger_path = Path(
+            f"/Users/alicedriessen/Box/CAR_Tcells/Model/conditional-monge/experiments/cmonge_ood/41BB_OOD/{data}/logs.yml"
+        )
     logger.info(f"Experiment: Leaving {config.ood_condition.conditions} out")
 
     datamodule = ConditionalDataModule(config.data, config.condition, config.ae)