benchmarking and number of tempaltes

Author: AmirAsgary

AFfine/predict_utils.py

@@ -398,7 +398,7 @@ def create_single_template_features(
     num_res_tmp = len(crs_tmp)
     template_full_sequence = ''.join(all_name1s_tmp[c][r] for c,r in crs_tmp)
     if expected_template_len:
-        print(expected_template_len, template_full_sequence, num_res)
+        #print(expected_template_len, template_full_sequence, num_res)
         assert len(template_full_sequence) == expected_template_len, (f'template name or row: {template_name} | '
                                                                       f'template_full_sequence: {template_full_sequence} |'
                                                                       f'expected_template_len: {expected_template_len}')

AFfine/run_prediction.py

@@ -106,7 +106,6 @@
           targets.shape[0], 'max_len=', crop_size)
 
 sys.stdout.flush()
-print('###########DEBUG', args.model_names)
 model_runners = predict_utils.load_model_runners(
     args.model_names,
     crop_size,
@@ -161,6 +160,7 @@
         except:
             data = data.sample(frac=1).reset_index(drop=True)
             data = data.iloc[:int(num_templates)]
+    print(f"AlphaFold long: Using {len(data)} templates")
 
 
     cols = ('template_pdbfile target_to_template_alignstring identities '

run_PMGen.py

@@ -140,7 +140,7 @@ def main():
 
     #benchamark flags
     parser.add_argument('--benchmark', action='store_true', help='Enable benchmarking')
-    parser.add_argument('--benchmark_similarity_threshold', type=float, default=1.0,
+    parser.add_argument('--benchmark_similarity_threshold', type=float, default=100.0,
         help='When --benchmark is set, exclude templates with peptide-MHC (pMHC) sequence '
             'similarity above this fraction (0-1). pMHC similarity is the length-weighted '
             'average of MHC G-domain identity and peptide identity. Default: 1.0.') # added after review --> similarity benchmark

run_utils.py

@@ -326,7 +326,8 @@ def run_alphafold(self, input_file, output_prefix):
             "--model_names", *model_names.split(),
             "--model_params_files", *model_params_files.split(),
             "--ignore_identities",
-            "--num_recycles", f"{self.num_recycles}"
+            "--num_recycles", f"{self.num_recycles}", 
+            "--num_templates", f"{self.num_templates}"
         ]
         if self.sampling_mode:
             command += ['--sampling_mode',
@@ -623,7 +624,8 @@ def run_wrapper_parallel(self, max_ram=3, max_cores=4, run_alphafold=True):
         runner = run_PMGen_modeling(peptide=row['peptide'], mhc_seq=row['mhc_seq'],
                                             mhc_type=row['mhc_type'], id=f"{row['id']}", output_dir=self.output_dir,
                                             anchors=anchors, mhc_allele=mhc_allele, predict_anchor=predict_anchor,
-                                            num_templates=self.num_templates, num_recycles=self.num_recycles,
+                                            num_templates=self.num_templates, best_n_templates=self.best_n_templates,
+                                            num_recycles=self.num_recycles,
                                             models=self.models, alphafold_param_folder=self.alphafold_param_folder,
                                             fine_tuned_model_path=self.fine_tuned_model_path, no_modelling=self.no_modelling,
                                             return_all_outputs=self.return_all_outputs,