draft seqfish

seqfish/to_zarr.py

@@ -0,0 +1,97 @@
+##
+import re
+import os
+import numpy as np
+import spatialdata as sd
+from napari_spatialdata import Interactive
+from dask_image.imread import imread
+import pandas as pd
+from anndata import AnnData
+import anndata as ad
+
+f = "data/Kidney Data 4 Download"
+
+##
+CELL_COORDINATES = "CellCoordinates"
+SECTION = "Section"
+CXG = "CxG"
+DAPI = "DAPI"
+
+
+def find_library_id_and_sections(data_root):
+    r_cell_coordinates = rf"^([A-Za-z0-9_]*?)_{CELL_COORDINATES}_{SECTION}([0-9]+).csv$"
+    files = os.listdir(data_root)
+    patterns = set()
+    sections = set()
+    for file in files:
+        m = re.match(r_cell_coordinates, file)
+        if m:
+            patterns.add(m.group(1))
+            sections.add(int(m.group(2)))
+    assert len(patterns) == 1
+    return patterns.pop(), sections
+
+
+def seqfish(data_root):
+    library_id, sections = find_library_id_and_sections(data_root)
+    # sections.remove(2)
+    # sections.remove(3)
+    print(library_id, sections)
+    images = {}
+    shapes = {}
+    tables = {}
+
+    for section in sections:
+        cell_coordinates_file = (
+            f"{library_id}_{CELL_COORDINATES}_{SECTION}{section}.csv"
+        )
+        cxg_file = f"{library_id}_{CXG}_{SECTION}{section}.csv"
+        image_file = f"{library_id}_{DAPI}_{SECTION}{section}.tiff"
+        assert os.path.isfile(os.path.join(data_root, cell_coordinates_file))
+        assert os.path.isfile(os.path.join(data_root, cxg_file))
+        assert os.path.isfile(os.path.join(data_root, image_file))
+        coords = pd.read_csv(os.path.join(data_root, cell_coordinates_file))
+        xy = coords[["center_x", "center_y"]].values
+        radius = np.sqrt(coords["area"] / np.pi)
+        label = coords["label"].values
+        cxg = pd.read_csv(os.path.join(data_root, cxg_file))
+        im = imread(os.path.join(data_root, image_file))
+        images[f"image_Section{section}"] = sd.models.Image2DModel.parse(
+            im,
+            dims=["c", "y", "x"],
+            scale_factors=[2, 2, 2, 2],
+            transformations={f"Section{section}": sd.transformations.Identity()},
+        )
+        shapes[f"cells_Section{section}"] = sd.models.ShapesModel.parse(
+            xy,
+            geometry=0,
+            radius=radius,
+            transformations={f"Section{section}": sd.transformations.Identity()},
+            index=label,
+        )
+        table = AnnData(
+            cxg.iloc[:, 1:],
+            obs=pd.DataFrame(
+                {"region": f"cells_Section{section}", "instance_id": cxg.iloc[:, 0]}
+            ),
+        )
+        tables[f"Section{section}"] = table
+
+    merged = ad.concat(tables)
+    table = sd.models.TableModel.parse(
+        merged,
+        region=[f"cells_Section{section}" for section in sections],
+        region_key="region",
+        instance_key="instance_id",
+    )
+
+    return sd.SpatialData(images=images, shapes=shapes, table=table)
+
+
+if __name__ == "__main__":
+    sdata = seqfish(f)
+    print(sdata)
+    sdata.write('data.zarr')
+    # interactive = Interactive(sdata)
+    # interactive.run()
+