Spaces:
Running
Running
simonduerr
commited on
Commit
•
ff1a3bf
1
Parent(s):
5bb015b
Update inference_app.py
Browse files- inference_app.py +30 -18
inference_app.py
CHANGED
@@ -87,7 +87,28 @@ def run_smina(
|
|
87 |
)
|
88 |
return output_text
|
89 |
|
90 |
-
def predict (input_sequence, input_ligand, input_protein):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
91 |
start_time = time.time()
|
92 |
|
93 |
if input_protein==None:
|
@@ -102,6 +123,7 @@ def predict (input_sequence, input_ligand, input_protein):
|
|
102 |
|
103 |
os.system(f"obabel {input_protein.name} -xr -O /usr/src/app/receptor.pdbqt")
|
104 |
os.system("obabel -isdf /usr/src/app/ligand.sdf -O /usr/src/app/ligand.pdbqt")
|
|
|
105 |
#Find pocket
|
106 |
pdb = md.load(input_protein.name)
|
107 |
# run ligsite
|
@@ -110,23 +132,15 @@ def predict (input_sequence, input_ligand, input_protein):
|
|
110 |
min_samples_value = 5
|
111 |
dbscan = DBSCAN(eps=eps_value, min_samples=min_samples_value)
|
112 |
labels = dbscan.fit_predict(pockets_xyz)
|
113 |
-
|
114 |
# Find the unique clusters and their sizes
|
115 |
unique_labels, counts = np.unique(labels, return_counts=True)
|
116 |
-
|
117 |
# Exclude noise points
|
118 |
valid_clusters = unique_labels[unique_labels != -1]
|
119 |
valid_counts = counts[unique_labels != -1]
|
120 |
-
|
121 |
# Find the cluster with the most points (highest density)
|
122 |
densest_cluster_label = valid_clusters[np.argmax(valid_counts)]
|
123 |
densest_cluster_points = pockets_xyz[labels == densest_cluster_label]
|
124 |
-
|
125 |
-
pocket_center = np.mean(densest_cluster_points, axis=0)
|
126 |
-
|
127 |
-
|
128 |
-
import pandas as pd
|
129 |
-
|
130 |
top_df = pd.DataFrame()
|
131 |
top_df['serial'] = list(range(densest_cluster_points.shape[0]))
|
132 |
top_df['name'] = 'PK'
|
@@ -134,24 +148,22 @@ def predict (input_sequence, input_ligand, input_protein):
|
|
134 |
top_df['resSeq'] = list(range(densest_cluster_points.shape[0]))
|
135 |
top_df['resName'] = 'PCK'
|
136 |
top_df['chainID'] = 0
|
137 |
-
|
138 |
pocket_top = md.Topology.from_dataframe(top_df, np.array([]))
|
139 |
pocket_trj = md.Trajectory(xyz=densest_cluster_points, topology=pocket_top)
|
140 |
pocket_trj.save('/usr/src/app/pockets_dense.pdb')
|
141 |
-
|
142 |
parser = PDBParser()
|
143 |
struc = parser.get_structure("X", "/usr/src/app/pockets_dense.pdb")
|
144 |
coords = [x.coord for x in struc.get_atoms()]
|
145 |
pocket_center = np.mean(coords, axis=0)
|
|
|
146 |
output_text = run_smina(
|
147 |
"/usr/src/app/ligand.pdbqt",
|
148 |
"/usr/src/app/receptor.pdbqt",
|
149 |
"/usr/src/app/docking_pose.sdf",
|
150 |
pocket_center,
|
151 |
[10,10,10],
|
152 |
-
|
153 |
-
|
154 |
-
|
155 |
end_time = time.time()
|
156 |
run_time = end_time - start_time
|
157 |
return [input_protein.name,"/usr/src/app/docking_pose.sdf"], run_time
|
@@ -170,7 +182,7 @@ with gr.Blocks() as app:
|
|
170 |
# define any options here
|
171 |
|
172 |
# for automated inference the default options are used
|
173 |
-
|
174 |
# checkbox_option = gr.Checkbox(label="Checkbox Option")
|
175 |
# dropdown_option = gr.Dropdown(["Option 1", "Option 2", "Option 3"], label="Radio Option")
|
176 |
|
@@ -181,7 +193,7 @@ with gr.Blocks() as app:
|
|
181 |
[
|
182 |
"SVKSEYAEAAAVGQEAVAVFNTMKAAFQNGDKEAVAQYLARLASLYTRHEELLNRILEKARREGNKEAVTLMNEFTATFQTGKSIFNAMVAAFKNGDDDSFESYLQALEKVTAKGETLADQIAKAL:SVKSEYAEAAAVGQEAVAVFNTMKAAFQNGDKEAVAQYLARLASLYTRHEELLNRILEKARREGNKEAVTLMNEFTATFQTGKSIFNAMVAAFKNGDDDSFESYLQALEKVTAKGETLADQIAKAL",
|
183 |
"COc1ccc(cc1)n2c3c(c(n2)C(=O)N)CCN(C3=O)c4ccc(cc4)N5CCCCC5=O",
|
184 |
-
"
|
185 |
],
|
186 |
],
|
187 |
[input_sequence, input_ligand, input_protein],
|
@@ -215,6 +227,6 @@ with gr.Blocks() as app:
|
|
215 |
out = Molecule3D(reps=reps)
|
216 |
run_time = gr.Textbox(label="Runtime")
|
217 |
|
218 |
-
btn.click(predict, inputs=[input_sequence, input_ligand, input_protein], outputs=[out, run_time])
|
219 |
|
220 |
app.launch()
|
|
|
87 |
)
|
88 |
return output_text
|
89 |
|
90 |
+
def predict (input_sequence, input_ligand, input_protein, exhaustiveness):
|
91 |
+
"""
|
92 |
+
Main prediction function that calls ligsite and smina
|
93 |
+
|
94 |
+
Parameters
|
95 |
+
----------
|
96 |
+
input_sequence: str
|
97 |
+
monomer sequence
|
98 |
+
input_ligand: str
|
99 |
+
ligand as SMILES string
|
100 |
+
protein_path: gradio.File
|
101 |
+
Gradio file object to monomer protein structure as PDB
|
102 |
+
exhaustiveness: int
|
103 |
+
SMINA parameter
|
104 |
+
|
105 |
+
Returns
|
106 |
+
-------
|
107 |
+
output_structures: tuple
|
108 |
+
(output_protein, output_ligand_sdf)
|
109 |
+
run_time: float
|
110 |
+
run time of the program
|
111 |
+
"""
|
112 |
start_time = time.time()
|
113 |
|
114 |
if input_protein==None:
|
|
|
123 |
|
124 |
os.system(f"obabel {input_protein.name} -xr -O /usr/src/app/receptor.pdbqt")
|
125 |
os.system("obabel -isdf /usr/src/app/ligand.sdf -O /usr/src/app/ligand.pdbqt")
|
126 |
+
|
127 |
#Find pocket
|
128 |
pdb = md.load(input_protein.name)
|
129 |
# run ligsite
|
|
|
132 |
min_samples_value = 5
|
133 |
dbscan = DBSCAN(eps=eps_value, min_samples=min_samples_value)
|
134 |
labels = dbscan.fit_predict(pockets_xyz)
|
|
|
135 |
# Find the unique clusters and their sizes
|
136 |
unique_labels, counts = np.unique(labels, return_counts=True)
|
|
|
137 |
# Exclude noise points
|
138 |
valid_clusters = unique_labels[unique_labels != -1]
|
139 |
valid_counts = counts[unique_labels != -1]
|
|
|
140 |
# Find the cluster with the most points (highest density)
|
141 |
densest_cluster_label = valid_clusters[np.argmax(valid_counts)]
|
142 |
densest_cluster_points = pockets_xyz[labels == densest_cluster_label]
|
143 |
+
# write cluster to PDB
|
|
|
|
|
|
|
|
|
|
|
144 |
top_df = pd.DataFrame()
|
145 |
top_df['serial'] = list(range(densest_cluster_points.shape[0]))
|
146 |
top_df['name'] = 'PK'
|
|
|
148 |
top_df['resSeq'] = list(range(densest_cluster_points.shape[0]))
|
149 |
top_df['resName'] = 'PCK'
|
150 |
top_df['chainID'] = 0
|
|
|
151 |
pocket_top = md.Topology.from_dataframe(top_df, np.array([]))
|
152 |
pocket_trj = md.Trajectory(xyz=densest_cluster_points, topology=pocket_top)
|
153 |
pocket_trj.save('/usr/src/app/pockets_dense.pdb')
|
|
|
154 |
parser = PDBParser()
|
155 |
struc = parser.get_structure("X", "/usr/src/app/pockets_dense.pdb")
|
156 |
coords = [x.coord for x in struc.get_atoms()]
|
157 |
pocket_center = np.mean(coords, axis=0)
|
158 |
+
# run smina
|
159 |
output_text = run_smina(
|
160 |
"/usr/src/app/ligand.pdbqt",
|
161 |
"/usr/src/app/receptor.pdbqt",
|
162 |
"/usr/src/app/docking_pose.sdf",
|
163 |
pocket_center,
|
164 |
[10,10,10],
|
165 |
+
exhaustiveness=exhaustiveness
|
166 |
+
)
|
|
|
167 |
end_time = time.time()
|
168 |
run_time = end_time - start_time
|
169 |
return [input_protein.name,"/usr/src/app/docking_pose.sdf"], run_time
|
|
|
182 |
# define any options here
|
183 |
|
184 |
# for automated inference the default options are used
|
185 |
+
exhaustiveness = gr.Slider(1,10,value=1, label="Slider Option")
|
186 |
# checkbox_option = gr.Checkbox(label="Checkbox Option")
|
187 |
# dropdown_option = gr.Dropdown(["Option 1", "Option 2", "Option 3"], label="Radio Option")
|
188 |
|
|
|
193 |
[
|
194 |
"SVKSEYAEAAAVGQEAVAVFNTMKAAFQNGDKEAVAQYLARLASLYTRHEELLNRILEKARREGNKEAVTLMNEFTATFQTGKSIFNAMVAAFKNGDDDSFESYLQALEKVTAKGETLADQIAKAL:SVKSEYAEAAAVGQEAVAVFNTMKAAFQNGDKEAVAQYLARLASLYTRHEELLNRILEKARREGNKEAVTLMNEFTATFQTGKSIFNAMVAAFKNGDDDSFESYLQALEKVTAKGETLADQIAKAL",
|
195 |
"COc1ccc(cc1)n2c3c(c(n2)C(=O)N)CCN(C3=O)c4ccc(cc4)N5CCCCC5=O",
|
196 |
+
"input_test.pdb"
|
197 |
],
|
198 |
],
|
199 |
[input_sequence, input_ligand, input_protein],
|
|
|
227 |
out = Molecule3D(reps=reps)
|
228 |
run_time = gr.Textbox(label="Runtime")
|
229 |
|
230 |
+
btn.click(predict, inputs=[input_sequence, input_ligand, input_protein, exhaustiveness], outputs=[out, run_time])
|
231 |
|
232 |
app.launch()
|