Updated inference procedure of deepnano and guppy to increase their performance

compare_benchmark_performance/compare_accuracy.py

@@ -114,9 +114,18 @@ def parse_paf(paf_path, ground_truth):
     y_true = merged_df['species id'].apply(lambda x:
                                            1 if x.find(target_id) == 0
                                            else 0)
-    # Mapping to a * means mapping to no species
-    y_pred = merged_df['Target sequence name'].apply(
-        lambda x: 0 if x == '*' else 1)
+    if tool == 'deepnano':
+        residue_matches = np.array(merged_df["Number of residue matches"])
+        alignment_length = np.array(merged_df["Alignment block length"])
+        align_perc = residue_matches / alignment_length
+        y_pred = (align_perc > 0.50).astype(int)
+    elif tool == 'uncalled':
+        # Mapping to a * means mapping to no species
+        y_pred = merged_df['Target sequence name'].apply(
+            lambda x: 0 if x == '*' else 1)
+    else:
+        raise ValueError('Tool not found')
+
     f1 = f1_score(y_true, y_pred)
     accuracy = accuracy_score(y_true, y_pred)
     cm = confusion_matrix(y_true, y_pred)
@@ -204,8 +213,7 @@ def calculate_accuracy_from_output(args):
         "squigglenet/*/fold?/inference")
     # Prepare arguments to parse to the functions in the pool
     args_baseless = [[file, ground_truth] for file in baseless_files]
-    args_uncalled_deepnano = [[file, ground_truth] for file
-                              in chain(deepnano_files, uncalled_files)]
+    args_uncalled = [[file, ground_truth] for file in uncalled_files]
     args_guppy = [[file, ground_truth] for file in guppy_files]
     args_squigglenet = [[folder, ground_truth]
                         for folder in squigglenet_files]
@@ -213,11 +221,11 @@ def calculate_accuracy_from_output(args):
         baseless_results = p.starmap(parse_baseless_output, args_baseless)
         squigglenet_results = p.starmap(parse_squigglenet_output,
                                         args_squigglenet)
-        uncalled_deepnano_results = p.starmap(parse_paf,
-                                              args_uncalled_deepnano)
+        uncalled_results = p.starmap(parse_paf, args_uncalled)
         guppy_results = p.starmap(parse_sequencing_summary, args_guppy)
-    all_records = (uncalled_deepnano_results + guppy_results
-                   + squigglenet_results + baseless_results)
+    deepnano_results = [parse_paf(file, ground_truth) for file in deepnano_files]
+    all_records = (guppy_results + deepnano_results + uncalled_results
+                   + baseless_results + squigglenet_results)
     assert all_records, ('No files found: make sure the input folder '
                          'contains directories called "uncalled",'
                          ' "guppy", "baseless" and "deepnano"')