Minor bug fixes

app/main.py

@@ -1,3 +1,5 @@
+import io
+
 import spaces
 
 import os
@@ -427,81 +429,82 @@ with gr.Blocks(theme=THEME, title='GenFBDD', css=static.CSS, delete_cache=(3600,
                                 value='Extract Pocket', variant='primary', interactive=True
                             )
 
-                    with gr.Row():
-                        with gr.Column(variant='panel'):
-                            gr.Markdown('## Dock Phase Settings')
-                            dock_n_poses = gr.Slider(
-                                value=5, minimum=1, maximum=20, step=1,
-                                label="Number of conformers to generate per fragment",
-                                interactive=True
+                with gr.Row():
+                    with gr.Column(variant='panel'):
+                        gr.Markdown('## Dock Phase Settings')
+                        dock_n_poses = gr.Slider(
+                            value=5, minimum=1, maximum=20, step=1,
+                            label="Number of conformers to generate per fragment",
+                            interactive=True
+                        )
+                        dock_confidence_cutoff = gr.Slider(
+                            value=-1.0, minimum=-2.0, maximum=0, step=0.1,
+                            label="Confidence cutoff for filtering conformers of docked fragments (>0: high, <=-1.5: low)",
+                            interactive=True
+                        )
+                        with gr.Accordion(label='Advanced Options', open=False):
+                            dock_model = gr.Dropdown(
+                                label='Select a Fragment Docking Model',
+                                choices=['DiffDock-L'],
+                                interactive=True,
                             )
-                            dock_confidence_cutoff = gr.Slider(
-                                value=-1.0, minimum=-2.0, maximum=0, step=0.1,
-                                label="Confidence cutoff for filtering conformers of docked fragments (>0: high, <=-1.5: low)",
+                            dock_steps = gr.Slider(
+                                minimum=20, maximum=40, step=1,
+                                label="Number of Denoising Steps for Docking Fragments",
                                 interactive=True
                             )
-                            with gr.Accordion(label='Advanced Options', open=False):
-                                dock_model = gr.Dropdown(
-                                    label='Select a Fragment Docking Model',
-                                    choices=['DiffDock-L'],
-                                    interactive=True,
-                                )
-                                dock_steps = gr.Slider(
-                                    minimum=20, maximum=40, step=1,
-                                    label="Number of Denoising Steps for Docking Fragments",
-                                    interactive=True
-                                )
-                        with gr.Column(variant='panel'):
-                            gr.Markdown('## Link Phase Settings')
-                            link_frag_pose_strategy = gr.Radio(
-                                label='Select a Fragment-Conformer Linking Strategy',
-                                choices=[
-                                    'Link Pairs of Fragment-Conformers Contacting the Pocket',
-                                    # 'Link Maximal Fragment-Conformers Spanning the Entire Pocket',
-                                ],
-                                value='Link Pairs of Fragment-Conformers Contacting the Pocket',
+                    with gr.Column(variant='panel'):
+                        gr.Markdown('## Link Phase Settings')
+                        link_frag_pose_strategy = gr.Radio(
+                            label='Select a Fragment-Conformer Linking Strategy',
+                            choices=[
+                                'Link Pairs of Fragment-Conformers Contacting the Pocket',
+                                # 'Link Maximal Fragment-Conformers Spanning the Entire Pocket',
+                            ],
+                            value='Link Pairs of Fragment-Conformers Contacting the Pocket',
+                        )
+                        link_frag_dist_range = RangeSlider(
+                            value=[2, 8], minimum=1, maximum=10, step=1,
+                            label="Fragment-Conformer Distance Range (Å) Eligible for Linking",
+                            interactive=True
+                        )
+                        link_n_mols = gr.Slider(
+                            value=10, minimum=1, maximum=20, step=1,
+                            label="Number of molecules to generate per fragment conformer combination",
+                            interactive=True
+                        )
+                        with gr.Accordion(label='Advanced Options', open=False):
+                            link_model = gr.Dropdown(
+                                label='Select a Linker Generation Model',
+                                choices=['DiffLinker'],
+                                interactive=True,
                             )
-                            link_frag_dist_range = RangeSlider(
-                                value=[2, 8], minimum=1, maximum=10, step=1,
-                                label="Fragment-Conformer Distance Range (Å) Eligible for Linking",
+                            link_linker_size = gr.Slider(
+                                minimum=0, maximum=20, step=1,
+                                label="Linker Size",
+                                info="0: automatically predicted; >=1: fixed size",
                                 interactive=True
                             )
-                            link_n_mols = gr.Slider(
-                                value=10, minimum=1, maximum=20, step=1,
-                                label="Number of molecules to generate per fragment conformer combination",
+                            link_steps = gr.Slider(
+                                minimum=100, maximum=500, step=10,
+                                label="Number of Denoising Steps for Generating Linkers",
                                 interactive=True
                             )
-                            with gr.Accordion(label='Advanced Options', open=False):
-                                link_model = gr.Dropdown(
-                                    label='Select a Linker Generation Model',
-                                    choices=['DiffLinker'],
-                                    interactive=True,
-                                )
-                                link_linker_size = gr.Slider(
-                                    minimum=0, maximum=20, step=1,
-                                    label="Linker Size",
-                                    info="0: automatically predicted; >=1: fixed size",
-                                    interactive=True
-                                )
-                                link_steps = gr.Slider(
-                                    minimum=100, maximum=500, step=10,
-                                    label="Number of Denoising Steps for Generating Linkers",
-                                    interactive=True
-                                )
-                    with gr.Row(equal_height=True):
-                        email_input =gr.Textbox(
-                            label='Email Address (Optional)',
-                            info="Your email address will be used to notify you of the status of your job. "
-                                 "If you cannot receive the email, please check your spam/junk folder.",
-                            type='email'
+                with gr.Row(equal_height=True):
+                    email_input =gr.Textbox(
+                        label='Email Address (Optional)',
+                        info="Your email address will be used to notify you of the status of your job. "
+                             "If you cannot receive the email, please check your spam/junk folder.",
+                        type='email'
+                    )
+                    with gr.Column():
+                        start_clr_btn = gr.ClearButton(
+                            value='Reset Inputs', interactive=True,
                         )
-                        with gr.Column():
-                            start_clr_btn = gr.ClearButton(
-                                value='Reset Inputs', interactive=True,
-                            )
-                            run_btn = gr.Button(
-                                value='Run GenFBDD', variant='primary', interactive=True,
-                            )
+                        run_btn = gr.Button(
+                            value='Run GenFBDD', variant='primary', interactive=True,
+                        )
+
             with gr.Tab(label='Results', id='result'):
                 # Results
                 result_state = gr.State(value={})
@@ -787,6 +790,7 @@ with gr.Blocks(theme=THEME, title='GenFBDD', css=static.CSS, delete_cache=(3600,
     pred_lookup_example.click(
         fn=lambda: '80cf2658-7a1c-48d6-8372-61b978177fe6',
         outputs=[pred_lookup_id],
+        show_progress='hidden'
     ).success(
         fn=query_job_status,
         inputs=pred_lookup_id,
@@ -881,20 +885,24 @@ with gr.Blocks(theme=THEME, title='GenFBDD', css=static.CSS, delete_cache=(3600,
 
     def generate_result_zip(result_info, compound_mod_df, protein_file):
         result_path = Path(result_info['output_dir'])
-        filename = f'GenFBDD_{result_path.name}.zip'
-        zip_path = result_path / filename
+        zip_filename = f'GenFBDD_{result_path.name}.zip'
+        summary_filename = f'{result_info["type"]}_summary.csv'
+        zip_path = result_path / zip_filename
         cols_to_drop = ['mol', 'Compound', 'protein_path']
         compound_mod_df.drop(columns=[col for col in cols_to_drop if col in compound_mod_df.columns], inplace=True)
         compound_mod_df.rename(columns=fn.COL_ALIASES, inplace=True)
-        compound_mod_df.to_csv(result_path / f'{result_info["type"]}_summary.csv', index=False)
 
         with zipfile.ZipFile(zip_path, 'w') as zip_file:
             for file in result_path.rglob('*'):
-                if file.is_file() and file != zip_path:  # Skip directories and the zip file itself
+                # Skip directories, the zip file itself and the new summary file
+                if file.is_file() and file.name not in [zip_filename, summary_filename]:
                     archive_path = file.relative_to(result_path)
                     zip_file.write(file, arcname=archive_path)
             if Path(protein_file).name not in zip_file.namelist():
                 zip_file.write(Path(protein_file), arcname=Path(protein_file).name)
+            csv_buffer = io.BytesIO()
+            compound_mod_df.to_csv(csv_buffer, index=False)
+            zip_file.writestr(summary_filename, csv_buffer.getvalue())
 
         return gr.File(str(zip_path), visible=True)
 

inference.py

@@ -757,7 +757,7 @@ def select_fragment_pairs(
                         combined_mol = Chem.CombineMols(row1['ligand_mol'], row2['ligand_mol'])
                         complex_name = f"{protein_id}-{row1['ID1']}-{row2['ID1']}"
                         if 'ligand_conf_path' in row1 and 'ligand_conf_path' in row2:
-                            fragment_path = [row1['ligand_conf_path'], row2['ligand_conf_path']]
+                            fragment_path = [row1['ligand_conf_path'].name, row2['ligand_conf_path'].name]
                         protein_fragment_conf_pairs.append(
                             (complex_name, protein_path,  # pocket_path,
                              combined_smiles, fragment_path, combined_mol, dist)
@@ -978,7 +978,7 @@ def generate_linkers(
     if 'protein_path' not in df.columns:
         df['protein_path'] = df['X2']
     if 'name' not in df.columns and 'ID1' in df.columns and 'ID2' in df.columns:
-        df['name'] = df['ID2'] + '-' + df['ID1']
+        df['name'] = df['ID1'] + '-' + df['ID2']
     df.dropna(subset=['fragment_mol', 'protein_path'], inplace=True)
 
     cached_parse_molecule = cache(parse_molecule)

results/2025-01-01_01-01-01_80cf2658-7a1c-48d6-8372-61b978177fe6/linking_summary.csv

@@ -1,7 +1,7 @@
 name,fragment_path,X1,X1^,out_path
-1xkk-frag_6_0-frag_9_0,"[PosixPath('results/2025-01-05_11-14-29.222981/docking/1xkk-frag_6_0-confidence-0.98.sdf'), PosixPath('results/2025-01-05_11-14-29.222981/docking/1xkk-frag_9_0-confidence-1.02.sdf')]",C1CCNCC1.Oc1ccccc1O,[C][C](Oc1[c][c][c][c]c1[O])C(=O)[N][C][C]N1[C][C][C][C][C]1,1xkk-frag_6_0-frag_9_0_0.sdf
-1xkk-frag_6_0-frag_1_0,"[PosixPath('results/2025-01-05_11-14-29.222981/docking/1xkk-frag_6_0-confidence-0.98.sdf'), PosixPath('results/2025-01-05_11-14-29.222981/docking/1xkk-frag_1_0-confidence0.01.sdf')]",C1CCNCC1.c1ccc2ncncc2c1,[O][C][C]c1n[c]c2[c][c][c]c([C]3[C][C][C][N][C]3)c2n1,1xkk-frag_6_0-frag_1_0_0.sdf
-1xkk-frag_6_0-frag_7_0,"[PosixPath('results/2025-01-05_11-14-29.222981/docking/1xkk-frag_6_0-confidence-0.98.sdf'), PosixPath('results/2025-01-05_11-14-29.222981/docking/1xkk-frag_7_0-confidence0.25.sdf')]",C1CCNCC1.c1ncc2cc[nH]c2n1,[C][C][C]C1=N[C]c2[c][c]n([C]3[C][C][C][N][C]3)c2[N]1,1xkk-frag_6_0-frag_7_0_0.sdf
-1xkk-frag_3_0-frag_4_0,"[PosixPath('results/2025-01-05_11-14-29.222981/docking/1xkk-frag_3_0-confidence-1.00.sdf'), PosixPath('results/2025-01-05_11-14-29.222981/docking/1xkk-frag_4_0-confidence-0.21.sdf')]",C=CC=O.CN1CCNCC1,[C]N1[C][C][N][C][C]1[C]C#C/[C]=[C]/[C]/[C]=[C]\C(=O)[C]=[C],1xkk-frag_3_0-frag_4_0_0.sdf
-1xkk-frag_3_0-frag_9_0,"[PosixPath('results/2025-01-05_11-14-29.222981/docking/1xkk-frag_3_0-confidence-1.00.sdf'), PosixPath('results/2025-01-05_11-14-29.222981/docking/1xkk-frag_9_0-confidence-1.02.sdf')]",C=CC=O.Oc1ccccc1O,[C][C](C(=O)[C]=[C])N1[C][C][C](C2=N[C]=C([N]C(=O)[N]c3[c][c]c([O])c([O])[c]3)[C]2)O[C]1,1xkk-frag_3_0-frag_9_0_0.sdf
-1xkk-frag_3_0-frag_1_0,"[PosixPath('results/2025-01-05_11-14-29.222981/docking/1xkk-frag_3_0-confidence-1.00.sdf'), PosixPath('results/2025-01-05_11-14-29.222981/docking/1xkk-frag_1_0-confidence0.01.sdf')]",C=CC=O.c1ccc2ncncc2c1,O=[C][C][C]C#C[C][C]c1[c][c]c2[c]n[c]nc2[c]1,1xkk-frag_3_0-frag_1_0_0.sdf
+1xkk-frag_6_0-frag_9_0,"['results/2025-01-05_11-14-29.222981/docking/1xkk-frag_6_0-confidence-0.98.sdf', 'results/2025-01-05_11-14-29.222981/docking/1xkk-frag_9_0-confidence-1.02.sdf']",C1CCNCC1.Oc1ccccc1O,[C][C](Oc1[c][c][c][c]c1[O])C(=O)[N][C][C]N1[C][C][C][C][C]1,1xkk-frag_6_0-frag_9_0_0.sdf
+1xkk-frag_6_0-frag_1_0,"['results/2025-01-05_11-14-29.222981/docking/1xkk-frag_6_0-confidence-0.98.sdf', 'results/2025-01-05_11-14-29.222981/docking/1xkk-frag_1_0-confidence0.01.sdf']",C1CCNCC1.c1ccc2ncncc2c1,[O][C][C]c1n[c]c2[c][c][c]c([C]3[C][C][C][N][C]3)c2n1,1xkk-frag_6_0-frag_1_0_0.sdf
+1xkk-frag_6_0-frag_7_0,"['results/2025-01-05_11-14-29.222981/docking/1xkk-frag_6_0-confidence-0.98.sdf', 'results/2025-01-05_11-14-29.222981/docking/1xkk-frag_7_0-confidence0.25.sdf']",C1CCNCC1.c1ncc2cc[nH]c2n1,[C][C][C]C1=N[C]c2[c][c]n([C]3[C][C][C][N][C]3)c2[N]1,1xkk-frag_6_0-frag_7_0_0.sdf
+1xkk-frag_3_0-frag_4_0,"['results/2025-01-05_11-14-29.222981/docking/1xkk-frag_3_0-confidence-1.00.sdf', 'results/2025-01-05_11-14-29.222981/docking/1xkk-frag_4_0-confidence-0.21.sdf']",C=CC=O.CN1CCNCC1,[C]N1[C][C][N][C][C]1[C]C#C/[C]=[C]/[C]/[C]=[C]\C(=O)[C]=[C],1xkk-frag_3_0-frag_4_0_0.sdf
+1xkk-frag_3_0-frag_9_0,"['results/2025-01-05_11-14-29.222981/docking/1xkk-frag_3_0-confidence-1.00.sdf', 'results/2025-01-05_11-14-29.222981/docking/1xkk-frag_9_0-confidence-1.02.sdf']",C=CC=O.Oc1ccccc1O,[C][C](C(=O)[C]=[C])N1[C][C][C](C2=N[C]=C([N]C(=O)[N]c3[c][c]c([O])c([O])[c]3)[C]2)O[C]1,1xkk-frag_3_0-frag_9_0_0.sdf
+1xkk-frag_3_0-frag_1_0,"['results/2025-01-05_11-14-29.222981/docking/1xkk-frag_3_0-confidence-1.00.sdf', 'results/2025-01-05_11-14-29.222981/docking/1xkk-frag_1_0-confidence0.01.sdf']",C=CC=O.c1ccc2ncncc2c1,O=[C][C][C]C#C[C][C]c1[c][c]c2[c]n[c]nc2[c]1,1xkk-frag_3_0-frag_1_0_0.sdf