fix: allow read_10x_mtx to read numeric gene IDs (#3932)

Author: flying-sheep

docs/release-notes/3932.fix.md

@@ -0,0 +1 @@
+Allow {func}`scanpy.read_10x_mtx` to read numeric gene IDs {smaller}`P Angerer`

src/scanpy/readwrite.py

@@ -628,21 +628,21 @@ def _read_10x_mtx(
         sep="\t",
     )
     if var_names == "gene_symbols":
-        var_names_idx = pd.Index(genes[1].values)
+        var_names_idx = pd.Index(genes[1].array)
         if make_unique:
             var_names_idx = anndata.utils.make_index_unique(var_names_idx)
-        adata.var_names = var_names_idx
-        adata.var["gene_ids"] = genes[0].values
+        adata.var_names = var_names_idx.astype("str")
+        adata.var["gene_ids"] = genes[0].array
     elif var_names == "gene_ids":
-        adata.var_names = genes[0].values
-        adata.var["gene_symbols"] = genes[1].values
+        adata.var_names = genes[0].array.astype("str")
+        adata.var["gene_symbols"] = genes[1].array
     else:
         msg = "`var_names` needs to be 'gene_symbols' or 'gene_ids'"
         raise ValueError(msg)
     if not is_legacy:
-        adata.var["feature_types"] = genes[2].values
+        adata.var["feature_types"] = genes[2].array
     barcodes = pd.read_csv(path / f"{prefix}barcodes.tsv{suffix}", header=None)
-    adata.obs_names = barcodes[0].values
+    adata.obs_names = barcodes[0].array
     return adata
 
 

tests/_data/10x_data/int-ids/barcodes.tsv

@@ -0,0 +1,5 @@
+AAACCCAAGATTAGCA-1
+AAACCCACACAATGAA-1
+AAACCCATCGGACCAC-1
+AAACCCATCTCGTTTA-1
+AAACGAAAGCAAGCCA-1

tests/_data/10x_data/int-ids/features.tsv

@@ -0,0 +1,6 @@
+574405	Pwwp4b	Gene Expression
+574404	Pwwp4a	Gene Expression
+574403	Insyn2b	Gene Expression
+574402	Gpr17	Gene Expression
+100862323	Btbd35f22	Gene Expression
+100862329	Btbd35f21	Gene Expression

tests/_data/10x_data/int-ids/matrix.mtx

@@ -0,0 +1,6 @@
+%%MatrixMarket matrix coordinate real general
+%
+6 5 3
+4 3 6
+5 3 1
+6 3 1

tests/test_read_10x.py

@@ -2,14 +2,20 @@
 
 import shutil
 from pathlib import Path
+from typing import TYPE_CHECKING
 from unittest.mock import patch
 
 import h5py
 import numpy as np
+import pandas as pd
 import pytest
 
 import scanpy as sc
 
+if TYPE_CHECKING:
+    from typing import Literal
+
+
 ROOT = Path(__file__).parent
 ROOT = ROOT / "_data" / "10x_data"
 VISIUM_ROOT = Path(__file__).parent / "_data" / "visium_data"
@@ -28,15 +34,19 @@ def assert_anndata_equal(a1, a2):
         pytest.param(
             ROOT / "1.2.0" / "filtered_gene_bc_matrices" / "hg19_chr21",
             ROOT / "1.2.0" / "filtered_gene_bc_matrices_h5.h5",
+            id="1.2.0",
         ),
         pytest.param(
             ROOT / "3.0.0" / "filtered_feature_bc_matrix",
             ROOT / "3.0.0" / "filtered_feature_bc_matrix.h5",
+            id="3.0.0",
         ),
     ],
 )
-@pytest.mark.parametrize("prefix", [None, "prefix_"])
-def test_read_10x(tmp_path, mtx_path, h5_path, prefix):
+@pytest.mark.parametrize("prefix", [None, "prefix_"], ids=["no_prefix", "prefix"])
+def test_read_10x(
+    tmp_path: Path, mtx_path: Path, h5_path: Path, prefix: str | None
+) -> None:
     if prefix is not None:
         # Build files named "prefix_XXX.xxx" in a temporary directory.
         mtx_path_orig = mtx_path
@@ -66,6 +76,27 @@ def test_read_10x(tmp_path, mtx_path, h5_path, prefix):
     assert_anndata_equal(sc.read_h5ad(from_mtx_pth), sc.read_h5ad(from_h5_pth))
 
 
+@pytest.mark.parametrize(
+    ("genes", "col_dtypes"),
+    [
+        pytest.param("symbols", dict(gene_ids="int64"), id="symbols"),
+        pytest.param("ids", dict(gene_symbols="str"), id="ids"),
+    ],
+)
+def test_read_10x_mtx_int(
+    genes: Literal["symbols", "ids"], col_dtypes: dict[str, str]
+) -> None:
+    str_dt = "str" if pd.options.future.infer_string else "object"
+    col_dtypes = {k: str_dt if v == "str" else v for k, v in col_dtypes.items()}
+
+    adata = sc.read_10x_mtx(
+        ROOT / "int-ids", var_names=f"gene_{genes}", compressed=False
+    )
+
+    assert adata.var.index.dtype == str_dt
+    assert dict(adata.var.dtypes) == dict(feature_types=str_dt, **col_dtypes)
+
+
 def test_read_10x_h5_v1():
     spec_genome_v1 = sc.read_10x_h5(
         ROOT / "1.2.0" / "filtered_gene_bc_matrices_h5.h5",