From 6756a06ffcaab803e8a5b6cb09fb5b0e09888bd2 Mon Sep 17 00:00:00 2001
From: "Chang.Liu" <TsumiNa@users.noreply.github.com>
Date: Sat, 8 Jun 2019 19:57:09 +0900
Subject: [PATCH] Add contribtion (#125)

* add Frozen Featurizer descriptor
* add Mordred descriptor
---
 licences/{placeholder => .gitkeep}            |   0
 requirements.txt                              |   1 +
 .../descriptor/test_mordred.py                |  60 ++++++++++++++++++
 tests/foo/descriptor/test_foo.py              |   8 ++-
 travis/linux-win/py36.yml                     |   1 +
 travis/linux-win/py37.yml                     |   1 +
 travis/osx/py36.yml                           |   1 +
 travis/osx/py37.yml                           |   1 +
 xenonpy/contrib/.DS_Store                     | Bin 0 -> 6148 bytes
 xenonpy/contrib/README.md                     |   7 +-
 xenonpy/contrib/extend_descriptors/README.md  |  20 ++++++
 .../contrib/extend_descriptors/__init__.py    |   3 +
 .../extend_descriptors/descriptor/__init__.py |   6 ++
 .../frozen_featurizer_descriptor.py           |  45 +++++++++++++
 .../descriptor/mordred_descriptor.py          |  39 ++++++++++++
 xenonpy/contrib/foo/__init__.py               |   2 -
 xenonpy/contrib/foo/descriptor/__init__.py    |   5 ++
 xenonpy/contrib/foo/{ => descriptor}/foo.py   |   2 +-
 xenonpy/descriptor/compositions.py            |   1 +
 xenonpy/descriptor/fingerprint.py             |  29 ++++++---
 xenonpy/descriptor/frozen_featurizer.py       |   1 +
 xenonpy/descriptor/structure.py               |  23 +++++++
 22 files changed, 241 insertions(+), 15 deletions(-)
 rename licences/{placeholder => .gitkeep} (100%)
 create mode 100644 tests/extend_descriptors/descriptor/test_mordred.py
 create mode 100644 xenonpy/contrib/.DS_Store
 create mode 100644 xenonpy/contrib/extend_descriptors/README.md
 create mode 100644 xenonpy/contrib/extend_descriptors/__init__.py
 create mode 100644 xenonpy/contrib/extend_descriptors/descriptor/__init__.py
 create mode 100644 xenonpy/contrib/extend_descriptors/descriptor/frozen_featurizer_descriptor.py
 create mode 100644 xenonpy/contrib/extend_descriptors/descriptor/mordred_descriptor.py
 create mode 100644 xenonpy/contrib/foo/descriptor/__init__.py
 rename xenonpy/contrib/foo/{ => descriptor}/foo.py (73%)

diff --git a/licences/placeholder b/licences/.gitkeep
similarity index 100%
rename from licences/placeholder
rename to licences/.gitkeep
diff --git a/requirements.txt b/requirements.txt
index cd20fb0d..587c0fea 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -3,6 +3,7 @@ tqdm
 seaborn
 plotly
 requests
+mordred
 
 ###### Requirements with Version Specifiers ######
 numpy == 1.16.*
diff --git a/tests/extend_descriptors/descriptor/test_mordred.py b/tests/extend_descriptors/descriptor/test_mordred.py
new file mode 100644
index 00000000..5d6cde2a
--- /dev/null
+++ b/tests/extend_descriptors/descriptor/test_mordred.py
@@ -0,0 +1,60 @@
+#  Copyright (c) 2019. TsumiNa. All rights reserved.
+#  Use of this source code is governed by a BSD-style
+#  license that can be found in the LICENSE file.
+
+import pandas as pd
+import pytest
+from mordred._base.pandas_module import MordredDataFrame
+from rdkit import Chem
+
+from xenonpy.contrib.extend_descriptors.descriptor import Mordred2DDescriptor
+
+
+@pytest.fixture(scope='module')
+def data():
+    # ignore numpy warning
+    import warnings
+    print('ignore NumPy RuntimeWarning\n')
+    warnings.filterwarnings("ignore", message="numpy.dtype size changed")
+    warnings.filterwarnings("ignore", message="numpy.ndarray size changed")
+
+    smis = ['C(C(O)C1(O))C(CO)OC1O',
+            'CC(C1=CC=CC=C1)CC(C2=CC=CC=C2)CC(C3=CC=CC=C3)CC(C4=CC=CC=C4)',
+            ' CC(C)CC(C)CC(C)',
+            'C(F)C(F)(F)']
+
+    mols = [Chem.MolFromSmiles(s) for s in smis]
+
+    err_smis = ['C(C(O)C1(O))C(CO)OC1O',
+                'CC(C1=CC=CC=C1)CC(C2=CC=CC=C2)CC(C3=CC=',
+                'Ccccccc',
+                'C(F)C(F)(F)']
+    yield dict(smis=smis, mols=mols, err_smis=err_smis)
+
+    print('test over')
+
+
+def test_mordred_1(data):
+    mordred = Mordred2DDescriptor()
+    desc = mordred.transform(data['smis'])
+    assert isinstance(desc, MordredDataFrame)
+
+    mordred = Mordred2DDescriptor(return_type='df')
+    desc = mordred.transform(data['smis'])
+    assert isinstance(desc, pd.DataFrame)
+
+
+def test_mordred_2(data):
+    mordred = Mordred2DDescriptor()
+    desc = mordred.transform(data['mols'])
+    assert isinstance(desc, MordredDataFrame)
+
+
+def test_mordred_3(data):
+    mordred = Mordred2DDescriptor()
+    with pytest.raises(ValueError):
+        mordred.transform(data['err_smis'])
+
+
+if __name__ == "__main__":
+    pytest.main()
diff --git a/tests/foo/descriptor/test_foo.py b/tests/foo/descriptor/test_foo.py
index 3a388599..693d6838 100644
--- a/tests/foo/descriptor/test_foo.py
+++ b/tests/foo/descriptor/test_foo.py
@@ -2,8 +2,14 @@
 #  Use of this source code is governed by a BSD-style
 #  license that can be found in the LICENSE file.
 
-from xenonpy.contrib.foo import hello_contrib
+import pytest
+
+from xenonpy.contrib.foo.descriptor import hello_contrib
 
 
 def test_foo_1():
     assert hello_contrib() == 'Hello contribution!'
+
+
+if __name__ == "__main__":
+    pytest.main()
diff --git a/travis/linux-win/py36.yml b/travis/linux-win/py36.yml
index e4b3af0d..8d630c7f 100644
--- a/travis/linux-win/py36.yml
+++ b/travis/linux-win/py36.yml
@@ -18,5 +18,6 @@ dependencies:
   - pip
   - pip:
       - ruamel.yaml
+      - mordred
       - pymatgen==2019.5.8
       - tqdm
diff --git a/travis/linux-win/py37.yml b/travis/linux-win/py37.yml
index 48707e52..33d29396 100644
--- a/travis/linux-win/py37.yml
+++ b/travis/linux-win/py37.yml
@@ -18,5 +18,6 @@ dependencies:
   - pip
   - pip:
       - ruamel.yaml
+      - mordred
       - pymatgen==2019.5.8
       - tqdm
diff --git a/travis/osx/py36.yml b/travis/osx/py36.yml
index 55efae91..0ee2139e 100644
--- a/travis/osx/py36.yml
+++ b/travis/osx/py36.yml
@@ -18,5 +18,6 @@ dependencies:
   - pip
   - pip:
       - ruamel.yaml
+      - mordred
       - pymatgen==2019.5.8
       - tqdm
diff --git a/travis/osx/py37.yml b/travis/osx/py37.yml
index 21d512df..4031b00b 100644
--- a/travis/osx/py37.yml
+++ b/travis/osx/py37.yml
@@ -18,5 +18,6 @@ dependencies:
   - pip
   - pip:
       - ruamel.yaml
+      - mordred
       - pymatgen==2019.5.8
       - tqdm
diff --git a/xenonpy/contrib/.DS_Store b/xenonpy/contrib/.DS_Store
new file mode 100644
index 0000000000000000000000000000000000000000..36e759217000a43b502425b55ea1ae8847a828bc
GIT binary patch
literal 6148
zcmeHKu};H44E2=`MJ#0j#B{&VKZvTp#?+AoltfU8kSH_@Q+9rUpWr{3*_ikhp3hcD
zl_G@-Aq2K$zl-mDsh_2|CL%Yl(;3l}h$c|R(Jsse;eFPgjMT%MF5HpPEuB(Em$d4g
zHpg#dfNOV*npITLJu3H6zZ$+}@KqFJ=Z=!#DN);ks++6XA}R8+lhjuC<V75(^(rBH
zJ$QJ0dE55)<VE@|_mtPYy!PpgGo}=p!46|%KvXC9!{?{^9JvpQYBT*p)P}q|rL(0m
zU<~{-2GFxvCVPTrjR9l87-$*b?}LXjMv9GK_;kR85db)bI|$Z1OK^-=j1(I|>_D7^
z0wvVxh~Xq0ey`^u#YRxV$?5Rn^vzBu6c^r(`+Yc0E)q0r3>X9J4D86`l>7h5=kx!1
zkUbd##=xIqz>VWYJV#2}TYJfIul3LiC<~u!1ltfSLMz5uZpBB?An<!$03*dl5FUvA
N5ePJxF$R8=fp;0tdV~M~

literal 0
HcmV?d00001

diff --git a/xenonpy/contrib/README.md b/xenonpy/contrib/README.md
index 8e18d3fc..beb9db29 100644
--- a/xenonpy/contrib/README.md
+++ b/xenonpy/contrib/README.md
@@ -9,10 +9,11 @@ get merged into XenonPy, but whose interfaces may still change, or which
 require some testing to see whether they can find broader acceptance.
 
 When adding a project, please stick to the following directory structure:
-Create a project directory in `contrib/`, and mirror the portions of the
-TensorFlow tree that your project requires underneath `contrib/my_project/`.
+1. Create a project directory in `contrib/`, and mirror the portions of the XenonPy tree that your project requires underneath `contrib/my_project/`.
+2. Provide a `README.md` under the root of the project directory, e.g  `contrib/my_project/README.md`.
 
-For example, let's say you create foo in `foo.py` and the testing codes
+
+For example, let's say you create a project named `foo` with source file `foo.py` and the testing file
 `foo_test.py`. If you were to merge those files directly into XenonPy,
 they would live in `$ROOT/xenonpy/descriptor/foo.py` and
 `$ROOT/tests/descriptor/foo_test.py`. In `contrib/`, they are part
diff --git a/xenonpy/contrib/extend_descriptors/README.md b/xenonpy/contrib/extend_descriptors/README.md
new file mode 100644
index 00000000..37ca835f
--- /dev/null
+++ b/xenonpy/contrib/extend_descriptors/README.md
@@ -0,0 +1,20 @@
+# Extend Descriptors
+
+## FrozenFeaturizerDescriptor
+
+This is a sample code for creating artificial descriptor based on a trained neural network.
+This code creates a BaseFeaturizer object in XenonPy that can be used as input for training models.
+The input is in the same format as the input of the descriptor used in the neural network.
+
+By passing both the XenonPy descriptor object and XenonPy frozen featurizer object into this class when creating the Base Featurizer, the output will be a dataframe same as other typical XenonPy descriptors, while the number of columns is the number of neurons in the chosen hidden layers.
+
+
+## Mordred2DDescriptor
+
+This is a sample code for calculating the 2D Mordred descriptor:
+https://github.com/mordred-descriptor/mordred
+
+This code creates a BaseFeaturizer object in XenonPy that can be used as input for training models.
+
+-----------
+written by Stephen Wu, 2019.05.31
diff --git a/xenonpy/contrib/extend_descriptors/__init__.py b/xenonpy/contrib/extend_descriptors/__init__.py
new file mode 100644
index 00000000..4a5d2d48
--- /dev/null
+++ b/xenonpy/contrib/extend_descriptors/__init__.py
@@ -0,0 +1,3 @@
+#  Copyright (c) 2019. TsumiNa. All rights reserved.
+#  Use of this source code is governed by a BSD-style
+#  license that can be found in the LICENSE file.
diff --git a/xenonpy/contrib/extend_descriptors/descriptor/__init__.py b/xenonpy/contrib/extend_descriptors/descriptor/__init__.py
new file mode 100644
index 00000000..b183f8d1
--- /dev/null
+++ b/xenonpy/contrib/extend_descriptors/descriptor/__init__.py
@@ -0,0 +1,6 @@
+#  Copyright (c) 2019. TsumiNa. All rights reserved.
+#  Use of this source code is governed by a BSD-style
+#  license that can be found in the LICENSE file.
+
+from .frozen_featurizer_descriptor import FrozenFeaturizerDescriptor
+from .mordred_descriptor import Mordred2DDescriptor
diff --git a/xenonpy/contrib/extend_descriptors/descriptor/frozen_featurizer_descriptor.py b/xenonpy/contrib/extend_descriptors/descriptor/frozen_featurizer_descriptor.py
new file mode 100644
index 00000000..1300d7a8
--- /dev/null
+++ b/xenonpy/contrib/extend_descriptors/descriptor/frozen_featurizer_descriptor.py
@@ -0,0 +1,45 @@
+#  Copyright (c) 2019. TsumiNa. All rights reserved.
+#  Use of this source code is governed by a BSD-style
+#  license that can be found in the LICENSE file.
+
+from typing import Union
+
+from xenonpy.descriptor import FrozenFeaturizer
+from xenonpy.descriptor.base import BaseFeaturizer, BaseDescriptor
+
+
+class FrozenFeaturizerDescriptor(BaseFeaturizer):
+
+    def __init__(self, descriptor_calculator: Union[BaseDescriptor, BaseFeaturizer],
+                 frozen_featurizer: FrozenFeaturizer, *,
+                 on_errors='raise',
+                 return_type='any'):
+        """
+        A featurizer for extracting artificial descriptors from neural networks
+
+        Parameters
+        ----------
+        descriptor_calculator : BaseFeaturizer or BaseDescriptor
+            Convert input data into descriptors to keep consistency with the pre-trained model.
+        frozen_featurizer : FrozenFeaturizer
+            Extracting artificial descriptors from neural networks
+        """
+
+        # fix n_jobs to be 0 to skip automatic wrapper in XenonPy BaseFeaturizer class
+        super().__init__(n_jobs=0, on_errors=on_errors, return_type=return_type)
+        self.FP = descriptor_calculator
+        self.ff = frozen_featurizer
+        self.output = None
+        self.__authors__ = ['Stephen Wu', 'TsumiNa']
+
+    def featurize(self, x, *, depth=1):
+        # transform input to descriptor dataframe
+        tmp_df = self.FP.transform(x)
+        # convert descriptor dataframe to hidden layer dataframe
+        self.output = self.ff.transform(tmp_df, depth=depth, return_type='df')
+        return self.output
+
+    @property
+    def feature_labels(self):
+        # column names based on xenonpy frozen featurizer setting
+        return self.output.columns
diff --git a/xenonpy/contrib/extend_descriptors/descriptor/mordred_descriptor.py b/xenonpy/contrib/extend_descriptors/descriptor/mordred_descriptor.py
new file mode 100644
index 00000000..210c1bdb
--- /dev/null
+++ b/xenonpy/contrib/extend_descriptors/descriptor/mordred_descriptor.py
@@ -0,0 +1,39 @@
+#  Copyright (c) 2019. TsumiNa. All rights reserved.
+#  Use of this source code is governed by a BSD-style
+#  license that can be found in the LICENSE file.
+
+from mordred import Calculator, descriptors
+from rdkit import Chem
+
+from xenonpy.descriptor.base import BaseFeaturizer
+
+
+class Mordred2DDescriptor(BaseFeaturizer):
+
+    def __init__(self, *, on_errors='raise', return_type='any'):
+        # fix n_jobs to be 0 to skip automatic wrapper in XenonPy BaseFeaturizer class
+        super().__init__(n_jobs=0, on_errors=on_errors, return_type=return_type)
+        self.output = None
+        self.__authors__ = ['Stephen Wu', 'TsumiNa']
+
+    def featurize(self, x):
+        # check if type(x) = list
+        if not isinstance(x, (list,)):
+            x = [x]
+        # check input format, assume SMILES if not RDKit-MOL
+        if not isinstance(x[0], Chem.rdchem.Mol):
+            x_mol = []
+            for z in x:
+                x_mol.append(Chem.MolFromSmiles(z))
+                if x_mol[-1] is None:
+                    raise ValueError('can not convert Mol from SMILES %s' % z)
+        else:
+            x_mol = x
+
+        calc = Calculator(descriptors, ignore_3D=True)
+        self.output = calc.pandas(x_mol)
+        return self.output
+
+    @property
+    def feature_labels(self):
+        return self.output.columns
diff --git a/xenonpy/contrib/foo/__init__.py b/xenonpy/contrib/foo/__init__.py
index d801c4f8..4a5d2d48 100644
--- a/xenonpy/contrib/foo/__init__.py
+++ b/xenonpy/contrib/foo/__init__.py
@@ -1,5 +1,3 @@
 #  Copyright (c) 2019. TsumiNa. All rights reserved.
 #  Use of this source code is governed by a BSD-style
 #  license that can be found in the LICENSE file.
-
-from .foo import hello_contrib
diff --git a/xenonpy/contrib/foo/descriptor/__init__.py b/xenonpy/contrib/foo/descriptor/__init__.py
new file mode 100644
index 00000000..609acf4b
--- /dev/null
+++ b/xenonpy/contrib/foo/descriptor/__init__.py
@@ -0,0 +1,5 @@
+#  Copyright (c) 2019. yoshida-lab. All rights reserved.
+#  Use of this source code is governed by a BSD-style
+#  license that can be found in the LICENSE file.
+
+from .foo import hello_contrib
diff --git a/xenonpy/contrib/foo/foo.py b/xenonpy/contrib/foo/descriptor/foo.py
similarity index 73%
rename from xenonpy/contrib/foo/foo.py
rename to xenonpy/contrib/foo/descriptor/foo.py
index b6ca1ca5..bfe05663 100644
--- a/xenonpy/contrib/foo/foo.py
+++ b/xenonpy/contrib/foo/descriptor/foo.py
@@ -1,4 +1,4 @@
-#  Copyright (c) 2019. TsumiNa. All rights reserved.
+#  Copyright (c) 2019. yoshida-lab. All rights reserved.
 #  Use of this source code is governed by a BSD-style
 #  license that can be found in the LICENSE file.
 
diff --git a/xenonpy/descriptor/compositions.py b/xenonpy/descriptor/compositions.py
index 1e141887..24d3c631 100644
--- a/xenonpy/descriptor/compositions.py
+++ b/xenonpy/descriptor/compositions.py
@@ -36,6 +36,7 @@ def __init__(self, *, one_hot_vec=False, n_jobs=-1, on_errors='raise', return_ty
         super().__init__(n_jobs=n_jobs, on_errors=on_errors, return_type=return_type)
         self.one_hot_vec = one_hot_vec
         self._elems = self._elements.index.tolist()
+        self.__authors__ = ['TsumiNa']
 
     def mix_function(self, elems, nums):
         vec = np.zeros(len(self._elems), dtype=np.int)
diff --git a/xenonpy/descriptor/fingerprint.py b/xenonpy/descriptor/fingerprint.py
index d0699712..25065e80 100644
--- a/xenonpy/descriptor/fingerprint.py
+++ b/xenonpy/descriptor/fingerprint.py
@@ -21,7 +21,7 @@ def __init__(self, n_jobs=-1, *, fp_size=2048, input_type='mol', on_errors='rais
         ----------
         n_jobs: int
             The number of jobs to run in parallel for both fit and predict.
-            Set -1 to use all cpu cores (default).
+            Can be -1 or # of cups. Set -1 to use all cpu cores (default).
         fp_size: int
             Fingerprint size.
         input_type: string
@@ -41,6 +41,7 @@ def __init__(self, n_jobs=-1, *, fp_size=2048, input_type='mol', on_errors='rais
         super().__init__(n_jobs=n_jobs, on_errors=on_errors, return_type=return_type)
         self.input_type = input_type
         self.fp_size = fp_size
+        self.__authors__ = ['Stephen Wu', 'TsumiNa']
 
     def featurize(self, x):
         if self.input_type == 'smiles':
@@ -77,7 +78,7 @@ def __init__(self, n_jobs=-1, *, n_bits=2048, input_type='mol', on_errors='raise
         ----------
         n_jobs: int
             The number of jobs to run in parallel for both fit and predict.
-            Set -1 to use all cpu cores (default).
+            Can be -1 or # of cups. Set -1 to use all cpu cores (default).
         n_bits: int
            Fixed bit length based on folding.
         input_type: string
@@ -97,6 +98,7 @@ def __init__(self, n_jobs=-1, *, n_bits=2048, input_type='mol', on_errors='raise
         super().__init__(n_jobs=n_jobs, on_errors=on_errors, return_type=return_type)
         self.input_type = input_type
         self.n_bits = n_bits
+        self.__authors__ = ['Stephen Wu', 'TsumiNa']
 
     def featurize(self, x):
         if self.input_type == 'smiles':
@@ -128,7 +130,8 @@ def __init__(self, n_jobs=-1, *, n_bits=2048, input_type='mol', on_errors='raise
         Parameters
         ----------
         n_jobs: int
-            The number of jobs to run in parallel for both fit and predict. Set -1 to use all cpu cores (default).
+            The number of jobs to run in parallel for both fit and predict.
+            Can be -1 or # of cups. Set -1 to use all cpu cores (default).
         n_bits: int
            Fixed bit length based on folding.
         input_type: string
@@ -148,6 +151,7 @@ def __init__(self, n_jobs=-1, *, n_bits=2048, input_type='mol', on_errors='raise
         super().__init__(n_jobs=n_jobs, on_errors=on_errors, return_type=return_type)
         self.input_type = input_type
         self.n_bits = n_bits
+        self.__authors__ = ['Stephen Wu', 'TsumiNa']
 
     def featurize(self, x):
         if self.input_type == 'smiles':
@@ -178,7 +182,8 @@ def __init__(self, n_jobs=-1, *, input_type='mol', on_errors='raise', return_typ
         Parameters
         ----------
         n_jobs: int
-            The number of jobs to run in parallel for both fit and predict. Set -1 to use all cpu cores (default).
+            The number of jobs to run in parallel for both fit and predict.
+            Can be -1 or # of cups. Set -1 to use all cpu cores (default).
         input_type: string
             Set the specific type of transform input.
             Set to ``mol`` (default) to ``rdkit.Chem.rdchem.Mol`` objects as input.
@@ -195,6 +200,7 @@ def __init__(self, n_jobs=-1, *, input_type='mol', on_errors='raise', return_typ
         """
         super().__init__(n_jobs=n_jobs, on_errors=on_errors, return_type=return_type)
         self.input_type = input_type
+        self.__authors__ = ['Stephen Wu', 'TsumiNa']
 
     def featurize(self, x):
         if self.input_type == 'smiles':
@@ -226,7 +232,8 @@ def __init__(self, n_jobs=-1, *, radius=3, n_bits=2048, input_type='mol', on_err
         Parameters
         ----------
         n_jobs: int
-            The number of jobs to run in parallel for both fit and predict. Set -1 to use all cpu cores (default).
+            The number of jobs to run in parallel for both fit and predict.
+            Can be -1 or # of cups. Set -1 to use all cpu cores (default).
         radius: int
             The radius parameter in the Morgan fingerprints, which is roughly half of the diameter parameter in FCFP,
             i.e., radius=2 is roughly equivalent to FCFP4.
@@ -250,6 +257,7 @@ def __init__(self, n_jobs=-1, *, radius=3, n_bits=2048, input_type='mol', on_err
         self.input_type = input_type
         self.radius = radius
         self.n_bits = n_bits
+        self.__authors__ = ['Stephen Wu', 'TsumiNa']
         # self.arg = arg # arg[0] = radius, arg[1] = bit length
 
     def featurize(self, x):
@@ -284,7 +292,8 @@ def __init__(self, n_jobs=-1, *, radius=3, n_bits=2048, input_type='mol', on_err
         Parameters
         ----------
         n_jobs: int
-            The number of jobs to run in parallel for both fit and predict. Set -1 to use all cpu cores (default).
+            The number of jobs to run in parallel for both fit and predict.
+            Can be -1 or # of cups. Set -1 to use all cpu cores (default).
         radius: int
             The radius parameter in the Morgan fingerprints, which is roughly half of the diameter parameter in ECFP,
             i.e., radius=2 is roughly equivalent to ECFP4.
@@ -308,6 +317,7 @@ def __init__(self, n_jobs=-1, *, radius=3, n_bits=2048, input_type='mol', on_err
         self.input_type = input_type
         self.radius = radius
         self.n_bits = n_bits
+        self.__authors__ = ['Stephen Wu', 'TsumiNa']
         # self.arg = arg # arg[0] = radius, arg[1] = bit length
 
     def featurize(self, x):
@@ -339,7 +349,8 @@ def __init__(self, n_jobs=-1, *, input_type='mol', on_errors='raise', return_typ
         Parameters
         ----------
         n_jobs: int
-            The number of jobs to run in parallel for both fit and predict. Set -1 to use all cpu cores (default).
+            The number of jobs to run in parallel for both fit and predict.
+            Can be -1 or # of cups. Set -1 to use all cpu cores (default).
         input_type: string
             Set the specific type of transform input.
             Set to ``mol`` (default) to ``rdkit.Chem.rdchem.Mol`` objects as input.
@@ -359,6 +370,7 @@ def __init__(self, n_jobs=-1, *, input_type='mol', on_errors='raise', return_typ
         self.input_type = input_type
         nms = [x[0] for x in Descriptors._descList]
         self.calc = MoleculeDescriptors.MolecularDescriptorCalculator(nms)
+        self.__authors__ = ['Stephen Wu', 'TsumiNa']
 
     def featurize(self, x):
         if self.input_type == 'smiles':
@@ -391,7 +403,8 @@ def __init__(self, n_jobs=-1, *, radius=3, n_bits=2048, fp_size=2048, input_type
         Parameters
         ----------
         n_jobs: int
-            The number of jobs to run in parallel for both fit and predict. Set -1 to use all cpu cores (default).
+            The number of jobs to run in parallel for both fit and predict.
+            Can be -1 or # of cups. Set -1 to use all cpu cores (default).
         radius: int
             The radius parameter in the Morgan fingerprints,
             which is roughly half of the diameter parameter in ECFP/FCFP,
diff --git a/xenonpy/descriptor/frozen_featurizer.py b/xenonpy/descriptor/frozen_featurizer.py
index f7e711ba..36e3d83d 100644
--- a/xenonpy/descriptor/frozen_featurizer.py
+++ b/xenonpy/descriptor/frozen_featurizer.py
@@ -45,6 +45,7 @@ def __init__(self, model=None, *, cuda=False, depth=None, on_errors='raise', ret
         self.model = model
         self.cuda = cuda
         self._ret = []
+        self.__authors__ = ['TsumiNa']
 
     def featurize(self, descriptor, *, depth=None):
         if not isinstance(self.model, tc.nn.Module):
diff --git a/xenonpy/descriptor/structure.py b/xenonpy/descriptor/structure.py
index e8a4b2ef..b249f13c 100644
--- a/xenonpy/descriptor/structure.py
+++ b/xenonpy/descriptor/structure.py
@@ -119,6 +119,29 @@ def __init__(self, including_d=True, *, n_jobs=-1, on_errors='raise', return_typ
         """
         super().__init__(n_jobs=n_jobs, on_errors=on_errors, return_type=return_type)
         self._including_d = including_d
+        self.__authors__ = ['TsumiNa']
+        self.__citations__ = [
+            '''
+            @article{LamPham2017,
+                archivePrefix = {arXiv},
+                arxivId = {1705.01043},
+                author = {{Lam Pham}, Tien and Kino, Hiori and Terakura, Kiyoyuki and Miyake, Takashi and Tsuda, Koji and Takigawa, Ichigaku and {Chi Dam}, Hieu},
+                doi = {10.1080/14686996.2017.1378060},
+                eprint = {1705.01043},
+                issn = {18785514},
+                journal = {Science and Technology of Advanced Materials},
+                keywords = {Material descriptor,data mining,machine learning,magnetic materials,material informatics},
+                number = {1},
+                pages = {756--765},
+                pmid = {29152012},
+                publisher = {Taylor {\&} Francis},
+                title = {{Machine learning reveals orbital interaction in materials}},
+                url = {https://doi.org/10.1080/14686996.2017.1378060},
+                volume = {18},
+                year = {2017}
+                }
+            '''
+        ]
 
     @staticmethod
     def get_element_representation(name):