Create app.py

app.py

@@ -0,0 +1,58 @@
+import gradio as gr
+import pandas as pd
+from rdkit import Chem
+from rdkit.Chem import AllChem
+import pickle
+
+# Load cell lines data and top genes
+cell_lines = pd.read_csv('gene_expression.csv', index_col=0)
+with open('2128_genes.pkl', 'rb') as f:
+    top_genes = pickle.load(f)
+
+# Load model
+with open('xgboost.pkl', 'rb') as f:
+    model = pickle.load(f)
+
+filtered_cell_lines = cell_lines[cell_lines.columns.intersection(top_genes)]
+
+# Define the smiles_to_fingerprint function
+def smiles_to_fingerprint(smiles):
+    mol = Chem.MolFromSmiles(smiles)
+    fp = AllChem.GetMorganFingerprintAsBitVect(mol, 2, nBits=1024)
+    return fp
+
+# Define a function that will be called when the user makes a prediction
+def predict(smiles_notation):
+    # Transform SMILES to fingerprint
+    fingerprint = smiles_to_fingerprint(smiles_notation)
+
+    # Convert the fingerprint to a DataFrame with one row and columns representing bits
+    fingerprint_df = pd.DataFrame([list(fingerprint)], columns=range(1024)).apply(lambda x: pd.Series({f'fp{str(i)}': val for i, val in enumerate(x)}), axis=1)
+
+    # Merge the fingerprint with each row of filtered_cell_lines
+    fingerprint_df['common_key'] = 1
+    filtered_cell_lines['common_key'] = 1
+    merged_data = pd.merge(filtered_cell_lines, fingerprint_df, on='common_key').drop('common_key', axis=1)
+
+    # Perform any additional processing or prediction based on the merged_data
+    predicts = model.predict(merged_data)
+
+    #merge predicts with cell lines
+    predicts = pd.DataFrame({'IC50': predicts,
+                             'Cell_line': filtered_cell_lines.index})
+
+    #sort by IC50 (only lowest 20)
+    predicts = predicts.sort_values(by='IC50').head(10)
+
+    return predicts
+
+# Define the Gradio interface
+iface = gr.Interface(
+    fn=predict,
+    inputs=gr.Textbox(value="COc1cc(O)c2c(c1)C=CCC(O)C(O)C(=O)C=CCC(C)OC2=O", lines=1, label="Enter drug in SMILES notation"),
+    outputs=gr.Dataframe(headers=['IC50', 'Cell_line'], type="numpy",label = 'Top 10 Cell Lines with lowest IC50 (GDSC2 dataset)' , datatype="number", row_count=10, col_count=2),
+    title="Drug Response Prediction",
+)
+
+# Launch the Gradio interface
+iface.launch(share=True)