cellucid 0.0.1a0__py3-none-any.whl

cellucid/__init__.py

@@ -0,0 +1,12 @@
+"""Utilities for exporting AnnData/array data for the 3D UMAP web viewer."""
+
+from importlib import metadata as _metadata
+
+try:
+    __version__ = _metadata.version("cellucid")
+except _metadata.PackageNotFoundError:  # pragma: no cover - handled in dev installs
+    __version__ = "0.0.0"
+
+from .prepare_data import export_data_for_web
+
+__all__ = ["export_data_for_web", "__version__"]

cellucid/anndata_variations.py

@@ -0,0 +1,149 @@
+import gc
+import pandas as pd
+import itertools
+from pathlib import Path
+from tqdm import tqdm
+
+try:
+    import scanpy as sc
+except ImportError as exc:  # pragma: no cover - optional dependency
+    raise ImportError(
+        "scanpy is required for cellucid.anndata_variations; install with "
+        "`pip install cellucid[analysis]` or add scanpy to your environment."
+    ) from exc
+
+def _find_repo_root() -> Path:
+    """Locate the python project root (contains pyproject.toml and data/)."""
+    for candidate in Path(__file__).resolve().parents:
+        if (candidate / "pyproject.toml").exists() and (candidate / "data").exists():
+            return candidate
+    return Path.cwd()
+
+
+REPO_ROOT = _find_repo_root()
+RAW_DATA_DIR = REPO_ROOT / "data" / "raw"
+EXPERIMENT_DIR = REPO_ROOT / "data" / "experiments"
+
+INPUT_H5AD = RAW_DATA_DIR / "adata_latent_scvi-adata_unified_20250925_001_integration-supergpu02.scidom.de-bb886d36409e78a3.h5ad"
+METADATA_PICKLE = RAW_DATA_DIR / "adata_obs_extracted.pickle"
+OUTPUT_DIR = EXPERIMENT_DIR / "umap_parameter_sweep"
+
+MIN_DISTS = [0.1, 0.15, 0.20, 0.25, 0.30, 0.35, 0.40, 0.45, 0.50]
+N_NEIGHBORS_LIST = [5, 10, 15, 30, 50, 90, 120]
+RANDOM_SEED = 42
+
+
+def _print_setup():
+    """Log resolved paths so it's obvious what is being used."""
+    print("==== AnnData UMAP sweep setup ====")
+    print(f"Current working dir : {Path.cwd()}")
+    print(f"Repo root           : {REPO_ROOT}")
+    print(f"Input h5ad          : {INPUT_H5AD} (exists: {INPUT_H5AD.exists()})")
+    print(f"Metadata pickle     : {METADATA_PICKLE} (exists: {METADATA_PICKLE.exists()})")
+    print(f"Output directory    : {OUTPUT_DIR}")
+    print("==================================")
+
+
+def validate_inputs() -> None:
+    """Ensure required files and folders exist before heavy work."""
+    missing = [path for path in (INPUT_H5AD, METADATA_PICKLE) if not path.exists()]
+    if missing:
+        missing_list = "\n - ".join(str(path) for path in missing)
+        raise FileNotFoundError(f"Missing required input files:\n - {missing_list}")
+
+    EXPERIMENT_DIR.mkdir(parents=True, exist_ok=True)
+    OUTPUT_DIR.mkdir(parents=True, exist_ok=True)
+
+
+
+def load_and_preprocess_data():
+    validate_inputs()
+    print("Loading and preparing data (this happens only once)...")
+    adata = sc.read_h5ad(INPUT_H5AD)
+    adata_complete = pd.read_pickle(METADATA_PICKLE)
+
+    if "qc_fail" not in adata_complete.columns:
+        raise KeyError("Metadata pickle missing required column 'qc_fail'.")
+
+    missing_cells = set(adata.obs_names) - set(adata_complete.index)
+    if missing_cells:
+        raise ValueError(f"Metadata missing {len(missing_cells)} cells (e.g., {next(iter(missing_cells))}).")
+
+    adata_complete = adata_complete.loc[adata.obs_names]
+    qc_mask = ~adata_complete["qc_fail"].fillna(False)
+
+    # Apply QC filtering to both the matrix and obs
+    adata = adata[qc_mask].copy()
+    adata_complete = adata_complete.loc[qc_mask]
+    adata.obs = adata_complete
+
+    # Drop heavy derived fields to reduce per-run memory
+    for container in (adata.obsm, adata.obsp):
+        for key in list(container.keys()):
+            del container[key]
+    adata.uns.clear()
+
+    print(f"Data prepared. Final: {adata}")
+    print(f"Cells kept after QC : {adata.n_obs}")
+    print(f"Features            : {adata.n_vars}")
+    return adata
+
+
+def run_parameter_sweep(_adata):
+    """
+    Iterates through parameters, calculating UMAPs and saving results.
+    """
+    OUTPUT_DIR.mkdir(parents=True, exist_ok=True)
+    param_combinations = list(itertools.product(N_NEIGHBORS_LIST, MIN_DISTS))
+    print(f"Starting sweep over {len(param_combinations)} combinations...")
+    print(f"Writing outputs to: {OUTPUT_DIR}")
+
+    existing_before = 0
+    for k, min_dist in param_combinations:
+        dist_str = f"{min_dist:.2f}".replace('.', '_')
+        filename = OUTPUT_DIR / f"adata|k__{k}|min_dist_{dist_str}.h5ad"
+        if filename.exists():
+            existing_before += 1
+    to_compute = len(param_combinations) - existing_before
+    print(f"Found {existing_before} outputs already present; {to_compute} will be computed.")
+
+    if to_compute == 0:
+        print("All combinations already exist. Nothing to do.")
+        return
+
+    skipped = 0
+    created = 0
+
+    for k, min_dist in tqdm(param_combinations, desc="Processing"):
+        dist_str = f"{min_dist:.2f}".replace('.', '_')
+        filename = OUTPUT_DIR / f"adata|k__{k}|min_dist_{dist_str}.h5ad"
+        if filename.exists():
+            skipped += 1
+            tqdm.write(f"Exists, skipping: {filename.name}")
+            continue
+
+        adata = _adata.copy()
+        try:
+            sc.pp.neighbors(adata, n_neighbors=k, random_state=RANDOM_SEED)
+            sc.tl.umap(
+                adata, 
+                n_components=3, 
+                min_dist=min_dist, 
+                random_state=RANDOM_SEED,
+            )
+            adata.write_h5ad(filename)
+            created += 1
+        except Exception as e:
+            tqdm.write(f"Calculation error for {filename.name}: {e}")
+        gc.collect()
+
+    print(f"Finished sweep. Created {created}, skipped {skipped}, total {len(param_combinations)}.")
+
+if __name__ == "__main__":
+    try:
+        _print_setup()
+        clean_adata = load_and_preprocess_data()
+        run_parameter_sweep(clean_adata)
+        print("Processing complete.")
+    except Exception as e:
+        print(f"An error occurred: {e}")