added plotMotif

concise/layers.py

@@ -4,9 +4,11 @@
 from keras import initializers
 from keras.layers.pooling import _GlobalPooling1D
 from keras.layers import Conv1D, Input
+from deeplift.visualization import viz_sequence
+import matplotlib.pyplot as plt
+
 from concise.regularizers import GAMRegularizer
 from concise.splines import BSpline
-import matplotlib.pyplot as plt
 
 
 # TODO - improve the naming
@@ -80,7 +82,7 @@ def __init__(self, filters,
         # override input shape
         if seq_length:
             kwargs["input_shape"] = (seq_length, 4)
-            kwargs["batch_input_shape"] = None
+            kwargs.pop("batch_input_shape", None)
 
         super(ConvDNA, self).__init__(
             filters=filters,
@@ -111,7 +113,41 @@ def get_config(self):
         config["seq_length"] = self.seq_length
         return config
 
-    # TODO - define the plotting function for motifs
+    # def plotWeights(self):
+    #     """Plot weights as matrices
+    #     """
+    #     pass
+
+    def plotMotif(self, index, figsize=(10, 2)):
+        """
+
+        Arguments:
+            index: Which motif to plot
+        """
+
+        W = self.get_weights()[0]
+
+        assert isinstance(index, int)
+        assert index >= 0
+        assert index < W.shape[2]
+        viz_sequence.plot_weights(W[:, :, index], figsize=figsize)
+
+    def plotMotifs(self, figsize=(10, 2)):
+        """
+
+        Arguments:
+            indices: Index list which ones to choose
+        """
+
+        W = self.get_weights()[0]
+
+        for index in range(W.shape[2]):
+            print("filter index: {0}".format(index))
+            viz_sequence.plot_weights(W[:, :, index], figsize=figsize)
+
+    # TODO - improve the plotting functions for motifs - refactor the viz_sequence
+    #        - mutliple panels with titles
+    #        - save to file if needed
 
 ############################################
 # Smoothing layers

nbs/01-simulated-data.ipynb

@@ -105,16 +105,24 @@
     "## Simulated data case study"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "In this notebook, we will replicate the results from [Plositional_effect/Simulation/01_fixed_seq_len.html](https://i12g-gagneurweb.in.tum.de/project/deepcis/#Scripts_Positional_effect_Simulation_01_fixed_seq_len.html) using concise models. Please have a look at that notebook first.\n",
+    "\n",
+    "\n"
+   ]
+  },
   {
    "cell_type": "code",
-   "execution_count": 37,
+   "execution_count": null,
    "metadata": {
     "collapsed": false
    },
    "outputs": [],
    "source": [
-    "## See dataset: \n",
-    "#%matplotlib notebook\n",
+    "# Used additional packages\n",
     "%matplotlib inline\n",
     "\n",
     "import matplotlib.pyplot as plt\n",
@@ -128,15 +136,6 @@
     "import keras.optimizers as ko"
    ]
   },
-  {
-   "cell_type": "markdown",
-   "metadata": {},
-   "source": [
-    "In this notebook, we will replicate the results from [Plositional_effect/Simulation/01_fixed_seq_len.html](https://i12g-gagneurweb.in.tum.de/project/deepcis/#Scripts_Positional_effect_Simulation_01_fixed_seq_len.html) using concise models. Please have a look at that notebook first.\n",
-    "\n",
-    "\n"
-   ]
-  },
   {
    "cell_type": "markdown",
    "metadata": {},

setup.py

@@ -22,6 +22,11 @@
     "tensorflow",
     "glmnet",
     "keras>=2.0.1",
+    'deeplift>=0.4',
+]
+
+dependency_links = [
+    "https://github.com/kundajelab/deeplift/tarball/v0.4.0-alpha#egg=deeplift-0.4"
 ]
 
 test_requirements = [
@@ -43,6 +48,7 @@
     include_package_data=True,
     setup_requires=['numpy'],
     install_requires=requirements,
+    dependency_links=dependency_links,
     license="MIT license",
     zip_safe=False,
     keywords=["computational biology", "bioinformatics", "genomics",

tests/test_conv1d.py

@@ -6,6 +6,7 @@
 import numpy as np
 from keras.utils.generic_utils import deserialize_keras_object, serialize_keras_object, get_custom_objects
 from keras.models import Sequential, model_from_json
+import matplotlib.pyplot as plt
 
 def test_correct_initialization():
     pwm_list = [PWM(np.array([[1, 2, 3, 4], [2, 4, 4, 5]])),
@@ -88,3 +89,37 @@ def test_init_serialization():
 
     # serialization was successfull
     assert np.all(a.kernel_initializer.pwm_list[0].pwm == pwm_list[0].pwm)
+
+
+def test_plot(tmpdir):
+    pwm_list = [PWM([[1, 2, 3, 4],
+                     [2, 4, 4, 5]]),
+                PWM([[1, 2, 1, 4],
+                     [2, 10, 4, 5]])]
+
+    # should work out of the box
+    # get_custom_objects()['PWMKernelInitializer'] = PWMKernelInitializer
+    # get_custom_objects()['PWMBiasInitializer'] = PWMBiasInitializer
+
+    seq_length = 100
+    input_shape = (None, seq_length, 4)  # (batch_size, steps, input_dim)
+    # input_shape = (seq_length, 4)  # (batch_size, steps, input_dim)
+
+    # output_shape = (None, steps, filters)
+    from concise.layers import ConvDNA
+    conv_l = ConvDNA(filters=15, kernel_size=11,
+                     kernel_regularizer=L1L2(l1=1, l2=1),  # Regularization
+                     activation="relu",
+                     kernel_initializer=PWMKernelInitializer(pwm_list, stddev=0.1),
+                     bias_initializer=PWMBiasInitializer(pwm_list, kernel_size=11),
+                     seq_length=seq_length
+                     )
+
+    conv_l.build(input_shape)
+
+    # plt.savefig('data.png')
+    # conv_l.plotMotif(0)
+    # plt.close()
+
+    # TODO
+