diff --git a/src/marshmallow/schema.py b/src/marshmallow/schema.py
index 92152696f..f2b3bf0e3 100644
--- a/src/marshmallow/schema.py
+++ b/src/marshmallow/schema.py
@@ -373,7 +373,7 @@ def __init__(
         context: dict | None = None,
         load_only: types.StrSequenceOrSet = (),
         dump_only: types.StrSequenceOrSet = (),
-        partial: bool | types.StrSequenceOrSet = False,
+        partial: bool | types.StrSequenceOrSet | None = None,
         unknown: str | None = None,
     ):
         # Raise error if only or exclude is passed as string, not list of strings
@@ -647,15 +647,18 @@ def _deserialize(
                         partial_is_collection and attr_name in partial
                     ):
                         continue
-                d_kwargs = {}
+                d_kwargs = {}  # type: typing.Dict[str, typing.Any]
                 # Allow partial loading of nested schemas.
                 if partial_is_collection:
-                    prefix = field_name + "."
-                    len_prefix = len(prefix)
-                    sub_partial = [
-                        f[len_prefix:] for f in partial if f.startswith(prefix)
-                    ]
-                    d_kwargs["partial"] = sub_partial
+                    if attr_name in partial:
+                        d_kwargs["partial"] = True
+                    else:
+                        prefix = f"{attr_name}."
+                        len_prefix = len(prefix)
+                        sub_partial = [
+                            f[len_prefix:] for f in partial if f.startswith(prefix)
+                        ]
+                        d_kwargs["partial"] = sub_partial
                 else:
                     d_kwargs["partial"] = partial
                 getter = lambda val: field_obj.deserialize(
@@ -1080,7 +1083,7 @@ def _invoke_load_processors(
         *,
         many: bool,
         original_data,
-        partial: bool | types.StrSequenceOrSet,
+        partial: bool | types.StrSequenceOrSet | None,
     ):
         # This has to invert the order of the dump processors, so run the pass_many
         # processors first.
@@ -1157,7 +1160,7 @@ def _invoke_schema_validators(
         data,
         original_data,
         many: bool,
-        partial: bool | types.StrSequenceOrSet,
+        partial: bool | types.StrSequenceOrSet | None,
         field_errors: bool = False,
     ):
         for attr_name in self._hooks[(VALIDATES_SCHEMA, pass_many)]:
diff --git a/tests/test_deserialization.py b/tests/test_deserialization.py
index 4ee5ba5b1..08d872edd 100644
--- a/tests/test_deserialization.py
+++ b/tests/test_deserialization.py
@@ -2019,17 +2019,18 @@ class Sch(Schema):
         assert "equal" in errors
         assert errors["equal"] == ["Must be equal to False."]
 
-    def test_nested_partial_load(self):
+    @pytest.mark.parametrize("partial", (True, {"z.x"}, {"z"}))
+    def test_nested_partial(self, partial):
         class SchemaA(Schema):
             x = fields.Integer(required=True)
-            y = fields.Integer()
+            y = fields.Integer(required=True)
 
         class SchemaB(Schema):
             z = fields.Nested(SchemaA)
 
         b_dict = {"z": {"y": 42}}
         # Partial loading shouldn't generate any errors.
-        result = SchemaB().load(b_dict, partial=True)
+        result = SchemaB().load(b_dict, partial=partial)
         assert result["z"]["y"] == 42
         # Non partial loading should complain about missing values.
         with pytest.raises(ValidationError) as excinfo:
@@ -2039,10 +2040,11 @@ class SchemaB(Schema):
         assert "z" in errors
         assert "x" in errors["z"]
 
-    def test_deeply_nested_partial_load(self):
+    @pytest.mark.parametrize("partial", (True, {"b.c.x"}, {"b.c"}, {"b"}))
+    def test_deeply_nested_partial(self, partial):
         class SchemaC(Schema):
             x = fields.Integer(required=True)
-            y = fields.Integer()
+            y = fields.Integer(required=True)
 
         class SchemaB(Schema):
             c = fields.Nested(SchemaC)
@@ -2052,7 +2054,7 @@ class SchemaA(Schema):
 
         a_dict = {"b": {"c": {"y": 42}}}
         # Partial loading shouldn't generate any errors.
-        result = SchemaA().load(a_dict, partial=True)
+        result = SchemaA().load(a_dict, partial=partial)
         assert result["b"]["c"]["y"] == 42
         # Non partial loading should complain about missing values.
         with pytest.raises(ValidationError) as excinfo:
@@ -2063,22 +2065,44 @@ class SchemaA(Schema):
         assert "c" in errors["b"]
         assert "x" in errors["b"]["c"]
 
-    def test_nested_partial_tuple(self):
+    @pytest.mark.parametrize("partial", (True, {"x"}))
+    def test_nested_partial_from_schema_init(self, partial):
         class SchemaA(Schema):
             x = fields.Integer(required=True)
             y = fields.Integer(required=True)
 
         class SchemaB(Schema):
-            z = fields.Nested(SchemaA)
+            z = fields.Nested(SchemaA(partial=partial))
 
         b_dict = {"z": {"y": 42}}
-        # If we ignore the missing z.x, z.y should still load.
+
+        result = SchemaB().load(b_dict)
+        assert "z" in result
+        assert "y" in result["z"]
+        assert "x" not in result["z"]
+        assert result["z"]["y"] == 42
+
+    def test_nested_with_data_keys_partial_using_tuple(self):
+        class SchemaA(Schema):
+            x = fields.Integer(required=True, data_key="x_key")
+            y = fields.Integer(required=True, data_key="y_key")
+
+        class SchemaB(Schema):
+            z = fields.Nested(SchemaA, data_key="z_key")
+
+        b_dict = {"z_key": {"y_key": 42}}
+
         result = SchemaB().load(b_dict, partial=("z.x",))
         assert result["z"]["y"] == 42
-        # If we ignore a missing z.y we should get a validation error.
-        with pytest.raises(ValidationError):
+
+        with pytest.raises(ValidationError) as exc:
             SchemaB().load(b_dict, partial=("z.y",))
 
+        data, errors = exc.value.valid_data, exc.value.messages
+        assert data["z"]["y"] == 42
+        assert "z_key" in errors
+        assert "x_key" in errors["z_key"]
+
 
 @pytest.mark.parametrize("FieldClass", ALL_FIELDS)
 def test_required_field_failure(FieldClass):  # noqa