Add a new checkcounts argument to skip raw count check

dca/__main__.py

@@ -114,6 +114,10 @@ def parse_args():
     parser.add_argument('--tensorboard', dest='tensorboard',
             action='store_true', help="Use tensorboard for saving weight distributions and "
                                       "visualization. (default: False)")
+    parser.add_argument('--checkcounts', dest='checkcounts', action='store_true',
+            help="Check if the expression matrix has raw (unnormalized) counts (default: True)")
+    parser.add_argument('--nocheckcounts', dest='checkcounts', action='store_false',
+            help="Do not check if the expression matrix has raw (unnormalized) counts")
     parser.add_argument('--denoisesubset', dest='denoisesubset', type=str,
                         help='Perform denoising only for the subset of genes '
                              'in the given file. Gene names should be line '
@@ -124,6 +128,7 @@ def parse_args():
                         saveweights=False,
                         sizefactors=True,
                         batchnorm=True,
+                        checkcounts=True,
                         norminput=True,
                         hyper=False,
                         debug=False,

dca/api.py

@@ -40,7 +40,8 @@ def dca(adata,
         training_kwds={},
         return_model=False,
         return_info=False,
-        copy=False
+        copy=False,
+        check_counts=True,
         ):
     """Deep count autoencoder(DCA) API.
 
@@ -116,6 +117,8 @@ def dca(adata,
         zinb or zinb-conddisp.
     copy : `bool`, optional. Default: `False`.
         If true, a copy of anndata is returned.
+    check_counts : `bool`. Default `True`.
+        Check if the counts are unnormalized (raw) counts.
 
     Returns
     -------
@@ -153,7 +156,8 @@ def dca(adata,
     adata = read_dataset(adata,
                          transpose=False,
                          test_split=False,
-                         copy=copy)
+                         copy=copy,
+                         check_counts=check_counts)
 
     # check for zero genes
     nonzero_genes, _ = sc.pp.filter_genes(adata.X, min_counts=1)

dca/io.py

@@ -50,7 +50,7 @@ def __getitem__(self, idx):
         return {'count': batch, 'size_factors': batch_sf}, batch
 
 
-def read_dataset(adata, transpose=False, test_split=False, copy=False):
+def read_dataset(adata, transpose=False, test_split=False, copy=False, check_counts=True):
 
     if isinstance(adata, sc.AnnData):
         if copy:
@@ -60,13 +60,14 @@ def read_dataset(adata, transpose=False, test_split=False, copy=False):
     else:
         raise NotImplementedError
 
-    # check if observations are unnormalized using first 10
-    X_subset = adata.X[:10]
-    norm_error = 'Make sure that the dataset (adata.X) contains unnormalized count data.'
-    if sp.sparse.issparse(X_subset):
-        assert (X_subset.astype(int) != X_subset).nnz == 0, norm_error
-    else:
-        assert np.all(X_subset.astype(int) == X_subset), norm_error
+    if check_counts:
+        # check if observations are unnormalized using first 10
+        X_subset = adata.X[:10]
+        norm_error = 'Make sure that the dataset (adata.X) contains unnormalized count data.'
+        if sp.sparse.issparse(X_subset):
+            assert (X_subset.astype(int) != X_subset).nnz == 0, norm_error
+        else:
+            assert np.all(X_subset.astype(int) == X_subset), norm_error
 
     if transpose: adata = adata.transpose()
 

dca/train.py

@@ -123,6 +123,7 @@ def train_with_args(args):
 
     adata = io.read_dataset(args.input,
                             transpose=(not args.transpose), # assume gene x cell by default
+                            check_counts=args.checkcounts,
                             test_split=args.testsplit)
 
     adata = io.normalize(adata,