adds test for zero genes to pca

docs/release-notes/0.10.0.md

@@ -15,6 +15,7 @@
 ```
 * Fixes an issue where `pp.normalize` and `pp.log1p` now use `copy` and `inplace` corretly {pr}`129` {smaller}`S Dicks`
 * changes the graph constructor for `tl.leiden` and `tl.louvain` {pr}`143` {smaller}`S Dicks`
+* Added a test to handle zero features, that caused issues in the sparse `pp.pca` {pr}`144` {smaller}`S Dicks`
 
 ```{rubric} Removals
 ```

src/rapids_singlecell/preprocessing/_kernels/_pca_sparse_kernel.py

@@ -1,5 +1,6 @@
 from __future__ import annotations
 
+import cupy as cp
 from cuml.common.kernel_utils import cuda_kernel_factory
 
 cov_kernel_str = r"""
@@ -50,6 +51,17 @@
     }
 }
 """
+check_zero_genes = r"""
+extern "C" __global__ void check_zero_genes(const int* indices, int* genes, int nnz) {
+    int value = blockIdx.x * blockDim.x + threadIdx.x;
+    if(value >= nnz){
+        return;
+    }
+    atomicAdd(&genes[indices[value]], 1);
+
+}
+"""
+_zero_genes_kernel = cp.RawKernel(check_zero_genes, "check_zero_genes")
 
 
 def _cov_kernel(dtype):

src/rapids_singlecell/preprocessing/_pca.py

@@ -210,6 +210,7 @@ def fit(self, x):
 
         if not isspmatrix_csr(x):
             x = x.tocsr()
+        _check_matrix_for_zero_genes(x)
         self.n_samples_ = x.shape[0]
         self.n_features_in_ = x.shape[1] if x.ndim == 2 else 1
         self.dtype = x.data.dtype
@@ -354,3 +355,26 @@ def _cov_sparse(x, return_gram=False, return_mean=False):
         return cov_result, mean_x, mean_x
     elif return_gram and return_mean:
         return cov_result, gram_matrix, mean_x, mean_x
+
+
+def _check_matrix_for_zero_genes(X):
+    gene_ex = cp.zeros(X.shape[1], dtype=cp.int32)
+
+    from ._kernels._pca_sparse_kernel import _zero_genes_kernel
+
+    block = (32,)
+    grid = (int(math.ceil(X.nnz / block[0])),)
+    _zero_genes_kernel(
+        grid,
+        block,
+        (
+            X.indices,
+            gene_ex,
+            X.nnz,
+        ),
+    )
+    if cp.any(gene_ex == 0):
+        raise ValueError(
+            "There are genes with zero expression. "
+            "Please remove them before running PCA."
+        )

src/rapids_singlecell/preprocessing/_utils.py

@@ -93,7 +93,8 @@ def _check_gpu_X(X):
     else:
         raise TypeError(
             "The input is not a CuPy ndarray or CuPy sparse matrix. "
-            "Rapids-singlecell only supports GPU matrices, "
+            "Rapids-singlecell only supports GPU matrices in this function, "
             "so your input must be either a CuPy ndarray or a CuPy sparse matrix. "
+            "Please checkout `rapids_singlecell.get.anndata_to_GPU` to convert your data to GPU. "
             "If you're working with CPU-based matrices, please consider using Scanpy instead."
         )

tests/test_pca.py

@@ -137,18 +137,6 @@ def test_pca_sparse():
         atol=1e-6,
     )
 
-def test_mask_highly_var_error():
-    """Check if use_highly_variable=True throws an error if the annotation is missing."""
-    adata = AnnData(np.array(A_list).astype("float32"))
-    with pytest.warns(
-        FutureWarning,
-        match=r"Argument `use_highly_variable` is deprecated, consider using the mask argument\.",
-    ), pytest.raises(
-        ValueError,
-        match=r"Did not find `adata\.var\['highly_variable'\]`\.",
-    ):
-        sc.pp.pca(adata, use_highly_variable=True)
-
 def test_mask_length_error():
     """Check error for n_obs / mask length mismatch."""
     adata = AnnData(np.array(A_list).astype("float32"))
@@ -244,3 +232,10 @@ def test_pca_layer():
     )
     np.testing.assert_almost_equal(X_adata.obsm["X_pca"], layer_adata.obsm["X_pca"])
     np.testing.assert_almost_equal(X_adata.varm["PCs"], layer_adata.varm["PCs"])
+
+def test_pca_layer_mask():
+    adata = sc.datasets.pbmc3k()[:,1000].copy()
+    sc.pp.normalize_total(adata)
+    sc.pp.log1p(adata)
+    with pytest.raises(ValueError,match="There are genes with zero expression. Please remove them before running PCA."):
+        rsc.pp.pca(adata)