prettier changelog, fixes in .network and .processing, tests

- prettier changelog - some style changes to improve Codacy - finetuning choose_correct_isoform for GlycoDraw formatting - fixing an issue in find_shared_virtuals - fixing an issue in distance_from_metrics - a few more tests
Glycocalex · Dec 19, 2024 · d2f5d55 · d2f5d55
1 parent dfb786d
commit d2f5d55
Show file tree

Hide file tree

Showing 8 changed files with 125 additions and 58 deletions.
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -2,49 +2,49 @@
 
 ## [1.5.0]
 
-### Added
+### Added ✨
 - Added type hints to all functions (e6721a1)
 - Added CodeCov shield to track PyTest test code coverage (23d6456)
 - Added more PyTest unit tests (e.g., 0c94995, 23d6456, 5a99d6b, f76535e, 94646ad, d5f5d4e, 918d18f, d1a8c6d, 194f31c)
 - Added setuptools to required_installs to support pip installation beyond `pip 25.0` (94646ad)
 - Added pyproject.toml to support pip installation beyond `pip 25.0` (94646ad)
 - Added CITATION.bib to allow for even easier citation of glycowork (a64f694)
 
-### Changed
+### Changed 🔄
 - Bumped minimum supported Python version to 3.9 (3.8 is no longer supported, see https://devguide.python.org/versions/) (4960c5c)
 - Switched docstring style to docments (<https://nbdev.fast.ai/tutorials/best_practices.html#document-parameters-with-docments>) (e6721a1)
 
 ### glycan_data
-##### Added
+##### Added ✨
 - Added new named motifs to `motif_list`: DisialylLewisC, "Sia(a2-3)Gal(b1-3)[Sia(a2-6)]GlcNAc"; RM2, "Sia(a2-3)[GalNAc(b1-4)]Gal(b1-3)[Sia(a2-6)]GlcNAc"; DisialylLewisA, "Sia(a2-3)Gal(b1-3)[Fuc(a1-4)][Sia(a2-6)]GlcNAc" (a64f694)
 - Added new curated glycomics dataset, `mouse_brain_GSL_PMID39375371` (b94744e)
 
-##### Changed
+##### Changed 🔄
 - Changed `glycoproteomics_human_keratinocytes_PMID37956981` to `glycoproteomics_human_keratinocytes_N_PMID37956981` (d5f5d4e)
 - Improved the description of blood group motifs in `motif_list` (including type 3 blood group antigens, ExtB, and parent motifs) (b94744e)
 
 #### loader
-##### Added
+##### Added ✨
 - Added `count_nested_brackets` helper function to monitor level of nesting in glycans (41bb1a1, d57b836)
 - Added dictionaries with lists of strings as values as a new supported data type for `DataFrameSerializer` (034b6ad)
 
-##### Changed
+##### Changed 🔄
 - Changed `resources.open_text` to `resources.files` to prevent `DeprecationWarning` from `importlib` (0c94995)
 - `lectin_specificity` now uses our custom `DataFrameSerializer` and is stored as a .json file rather than a .pkl file, to improve long-term stability across versions (034b6ad)
 
 #### stats
-##### Added
+##### Added ✨
 - Added the "random_state" keyword argument to `clr_transformation` to allow users to provide a reproducible RNG seed (b94744e)
 
-##### Changed
+##### Changed 🔄
 - For `replace_outliers_winsorization`, in small datasets, the 5% limit is dynamically changed to include at least one datapoint (23d6456)
 - Handled the edge case of strong differences in `cohen_d` with zero standard deviation; now outputting positive/negative infinity (23d6456)
 - Renamed `test_inter_vs_intra_group` to `compare_inter_vs_intra_group`, to avoid testing issues (23d6456)
 
-##### Deprecated
+##### Deprecated ⚠️
 - Deprecated `hlm`, `fast_two_sum`, `two_sum`, `expansion_sum`, and `update_cf_for_m_n`, which will all be done in-line instead (e1afe33)
 
-##### Fixed
+##### Fixed 🐛
 - Fixed DeprecationWarning in `calculate_permanova_stat` for calling nonzero on 0d arrays (23d6456)
 - Prevent possible division by zero in pseudo-F calculation in `calculate_permanova_stat` (23d6456)
 - Fixed DeprecationWarning in `jtkdist` for calling `np.sum(generator)` (23d6456)
@@ -55,11 +55,11 @@
 
 ### motif
 #### tokenization
-##### Added
+##### Added ✨
 - Added `get_random_glycan` to retrieve random glycan sequences (optionally of specific glycan type) (d1a8c6d)
 - Supported intramolecular modifications like lactonization in `glycan_to_composition` (8c69c2c)
 
-##### Changed
+##### Changed 🔄
 - Changed `resources.open_text` to `resources.files` to prevent `DeprecationWarning` from `importlib` (0c94995)
 - The monosaccharide keys of the output dictionaries of `glycan_to_composition` are now alphabetically sorted (8c69c2c)
 - Modified `calculate_adduct_mass` to deal with a greater variety of adduct handling, such as "C2H4O2", "-H2O", "+Na" to add or subtract masses (8c69c2c)
@@ -68,17 +68,17 @@
 - In addition to chemical formulae, users can now also provide direct additional masses as floats with the same "adduct" keyword argument in `composition_to_mass` and `glycan_to_mass` (d57b836)
 - `get_modification` will no longer return the 5Ac / 5Gc of Neu5Ac / Neu5Gc as part of the modification (0387d37)
 
-##### Fixed
+##### Fixed 🐛
 - Fixed an edge case in `get_unique_topologies`, in which the absence of a universal replacer sometimes created an empty list that was attempted to be indexed (0c94995)
 - Made sure that `compositions_to_structures` always returns a DataFrame, even if no matches are found (0c94995)
 
 #### processing
-##### Added
+##### Added ✨
 - Added "antennary_Fuc" as another inferred feature to `infer_features_from_composition` (a64f694)
 - Added "IdoA", "GalA", and "Araf" to recognized WURCS2 tokens (52fc16e)
 - Added the new "order_by" keyword argument to `choose_correct_isoform` to enforce strictly sorting branches by branch endings / linkages, if desired (918d18f)
 
-##### Changed
+##### Changed 🔄
 - `check_nomenclature` will now actually raise appropriate Exceptions, in case nomenclature is incompatible with glycowork, instead of print warnings (23d6456)
 - Supported triple-branch reordering in `find_isomorphs` and `choose_correct_isoform` (918d18f)
 - Improved `find_isomorphs` to swap neighboring branches with different levels of nesting (41bb1a1, 034b6ad)
@@ -94,24 +94,24 @@
 - Ensured that `canonicalize_iupac` works with lactonized glycans (i.e., containing something like "1,7lactone") (8c69c2c)
 - `find_matching_brackets_indices` has been renamed to `get_matching_indices` and now takes multiple delimiter choices and returns a generator, including the level of nesting (basically what `.draw.matches` used to do) (e1afe33)
 
-##### Fixed
+##### Fixed 🐛
 - Fixed component inference in `parse_glycoform` in case of unexpected composition formats (0c94995)
 - Fixed an issue in `equal_repeats`, in which identical repeats sometimes were not returning True (0c94995)
 
 #### graph
-##### Added
+##### Added ✨
 - Natively support narrow linkage ambiguity in `categorical_node_match_wildcard`; that means you can use things like "Gal(b1-3/4)GlcNAc" with `subgraph_isomorphism` or `compare_glycans` (as well as all functions using these core functions) and it will only return True for "Gal(b1-3)GlcNAc", "Gal(b1-4)GlcNAc", and "Gal(b1-?)GlcNAc" (b94744e)
 
-##### Changed
+##### Changed 🔄
 - Ensured that `compare_glycans` is 100% order-specific, never matching something like ("Gal(b1-4)GlcNAc", "GlcNAc(b1-4)Gal") (5a99d6b)
 - `glycan_to_nxGraph` will now return an empty graph if the input is an empty string (4f1ccfa)
 - `get_possible_topologies` will now also produce a warning (and return the input) if an already defined topology is provided as a pre-calculated graph (3f22f14)
 
 #### draw
-##### Added
+##### Added ✨
 - Added the "drawing" argument to `draw_hex`, `hex_circumference`, `add_bond`, `add_sugar`, and `draw_bracket` to avoid having to operate on global variables (918d18f)
 
-##### Changed
+##### Changed 🔄
 - `matches` can now also use [] as delimiters (f76535e)
 - Support easy import of `GlycoDraw`, via `from glycowork import GlycoDraw` (d5f5d4e)
 - Renamed `hex` to `draw_hex`, to avoid overwriting the built-in `hex` (918d18f)
@@ -121,20 +121,20 @@
 - Improved console drawing quality controlled by `display_svg_with_matplotlib` and image quality in Excel cells using `plot_glycans_excel` (a64f694)
 - `draw_chem2d` and `draw_chem3d` will now detect whether the user is in a Jupyter environment and, if not, plot to the Matplotlib console (c3a7f64)
 
-##### Deprecated
+##### Deprecated ⚠️
 - Deprecated `hex_circumference`, the functionality is now available within `draw_hex` with the new keyword argument "outline_only" (4f1ccfa)
 - Deprecated `multiple_branches`, `multiple_branch_branches`, `branch_order`, and `reorder_for_drawing` accordingly (41bb1a1)
 - Deprecated `matches`, which will now be done by `.processing.get_matching_indices` that has been reworked
 
-##### Fixed
+##### Fixed 🐛
 - Made sure `scale_in_range` never divides by zero, if value range is zero (f76535e)
 
 #### analysis
-##### Changed
+##### Changed 🔄
 - `get_glycanova` will now raise a ValueError if fewer than three groups are provided in the input data (f76535e)
 - Improved console drawing quality controlled by `display_svg_with_matplotlib` and image quality in Excel cells using `plot_glycans_excel` (a64f694)
 
-##### Fixed
+##### Fixed 🐛
 - Fixed a FutureWarning in `get_lectin_array` by avoiding DataFrame.groupby with axis=1 (f76535e)
 - Fixed a RuntimeWarning in `get_biodiversity` by handling statistical tests of identical alpha diversity values between groups (f76535e)
 - Made sure that the TSNE perplexity fits the sample size in `plot_embeddings` (d5f5d4e)
@@ -146,37 +146,41 @@
 - Ensured that "effect_size_variance = True" in `get_differential_expression` always formats variances correctly (ef3da9c)
 
 #### regex
-##### Fixed
+##### Fixed 🐛
 - Fixed an issue in `get_match_batch`, in which precompiled patterns caused issues in `get_match` (194f31c)
 
 #### annotate
-##### Added
+##### Added ✨
 - Added `get_size_branching_features` to create glycan size and branching level features for downstream analysis (d57b836)
 - Added the "size_branch" option in the "feature_set" keyword argument of `annotate_dataset` and `quantify_motifs`, to analyze glycans by size or level of branching (d57b836)
 
 ### ml
 #### model_training
-##### Added
+##### Added ✨
 - Added classification-AUROC, multilabel-accuracy, multilabel-MCC, regression-MAE, and regression-R2 as metrics to `train_model` (#66)
 - Added the "return_metrics" keyword argument to `train_model` that can additionally return all training and validation metrics (#66)
 
-##### Changed
+##### Changed 🔄
 - Weigh metric calculation by batch-size (correctly handling the last batch) in `train_model` (#66)
 - Best performances in `train_model` are now taken from the overall best model (lowest loss), not from best-model-per-metric (#66)
 
-##### Fixed
+##### Fixed 🐛
 - Fixed an indexing issue in `train_ml_model` if "additional_features_train" / "additional_features_test" were used (b94744e)
 
 #### inference
-##### Changed
+##### Changed 🔄
 - Changed `resources.open_text` to `resources.files` to prevent `DeprecationWarning` from `importlib` (d1a8c6d)
 
 ### network
 #### evolution
-##### Fixed
+##### Fixed 🐛
 - Fixed DeprecationWarning in `distance_from_embeddings` to prevent DataFrameGroupBy.apply from operating on the grouping columns (94646ad)
+- Fixed an issue in `distance_from_metric` where networks were indexed incorrectly based on presented DataFrame order
 
 #### biosynthesis
-##### Changed
+##### Changed 🔄
 - Made sure in `network_alignment` that only nodes that are virtual in all aligned networks stay virtual (918d18f)
 - `choose_leaves_to_extend` will now correctly return no leaf node glycan if the target composition cannot be reached from any of the leaf nodes in a network (918d18f)
+
+##### Fixed 🐛
+- Fixed an issue in `find_shared_virtuals` in which no shared nodes were found because of graph comparisons
diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md
@@ -35,19 +35,19 @@ pytest
 
 #### sub_module_name
 
-##### Added
+##### Added ✨
 
 * Added new feature X (commit-hash)
 
-##### Changed
+##### Changed 🔄
 
 * Modified behavior of Y (commit-hash)
 
-##### Deprecated
+##### Deprecated ⚠️
 
 * Removed feature Z (commit-hash)
 
-##### Fixed
+##### Fixed 🐛
 
 * Fixed bug in W (commit-hash)
 

diff --git a/glycowork/motif/graph.py b/glycowork/motif/graph.py
@@ -222,7 +222,7 @@ def expand_termini_list(motif: Union[str, nx.Graph], # Glycan motif sequence or
   num_linkages = motif.count('(') if isinstance(motif, str) else (len(motif) - 1) // 2
   result = [None] * (len(termini_list) + num_linkages)
   j = 0
-  for i in range(len(result)):
+  for i, _ in enumerate(result):
     if i % 2 == 0:
       result[i] = termini_list[j]
       j += 1
@@ -479,7 +479,7 @@ def try_string_conversion(graph: nx.Graph # Glycan graph to validate
     temp = graph_to_string(graph)
     temp = glycan_to_nxGraph(temp)
     return graph_to_string(temp)
-  except:
+  except (ValueError, IndexError):
     return None
 
 

diff --git a/glycowork/motif/processing.py b/glycowork/motif/processing.py
@@ -298,7 +298,7 @@ def compare_branches(branch1: str, branch2: str, use_linkage: bool = False) -> b
               kill_list.add(g)
   if order_by == "linkage":
     for g in glycans2:
-      if g[:g.index('[')].count('(') == 1 and g[g.index('['):g.index(']')].count('(') > 1:
+      if g[:g.index('[')].count('(') == 1 and g[g.index('['):g.index(']')].count('(') > 1 and g.count('[') > 1 and g.startswith('F'):
         kill_list.add(g)
       if pair_match := re.search(r'\[((?:[^[\]]|\[(?:[^[\]]|\[[^[\]]*\])*\])*)\]\[((?:[^[\]]|\[(?:[^[\]]|\[[^[\]]*\])*\])*)\]', g):
         if compare_branches(pair_match.group(1), pair_match.group(2)):

diff --git a/glycowork/motif/tokenization.py b/glycowork/motif/tokenization.py
@@ -156,7 +156,7 @@ def stemify_glycan(glycan: str, # Glycan in IUPAC-condensed format
           cut = glycan_start[index_pos:]
           try:
             cut = cut.split('(', 1)[0]
-          except:
+          except IndexError:
             pass
           # Replace offending monosaccharide with stemified monosaccharide
           if cut not in clean_list:

diff --git a/glycowork/network/biosynthesis.py b/glycowork/network/biosynthesis.py
@@ -147,8 +147,9 @@ def find_shared_virtuals(glycan_a: str, # First glycan
   ggraph_nb_b, _ = get_virtual_nodes(glycan_b, graph_dic, min_size = min_size)
   # Check whether any of the nodes of glycan_a and glycan_b are the same
   out = set((glycan_a, glycans_a[k]) for k, graph_a in enumerate(ggraph_nb_a)
-              for graph_b in ggraph_nb_b if compare_glycans(graph_a, graph_b))
-  out.update((glycan_b, g[1]) for g in out)
+              for graph_b in ggraph_nb_b if subgraph_isomorphism(graph_a, graph_b))
+  out2 = {(glycan_b, g[1]) for g in out}
+  out.update(out2)
   return list(out)
 
 
@@ -240,13 +241,10 @@ def find_shortest_path(goal_glycan: str, # Target glycan
   for glycan in sorted(glycan_list, key = len, reverse = True):
     # For each glycan, check whether it could constitute a precursor (i.e., is it a sub-graph + does it stem from the correct root)
     if len(glycan) < len(goal_glycan) and goal_glycan.endswith(glycan[-5:]) and subgraph_isomorphism(ggraph, safe_index(glycan, graph_dic)):
-      try:
-        # Finding a path through shells of generated virtual nodes
-        virtual_edges, edge_labels = find_path(goal_glycan, glycan, graph_dic,
-                                                 permitted_roots = permitted_roots, min_size = min_size, allowed_ptms = allowed_ptms)
-        return virtual_edges, edge_labels
-      except:
-        continue
+      # Finding a path through shells of generated virtual nodes
+      virtual_edges, edge_labels = find_path(goal_glycan, glycan, graph_dic,
+                                               permitted_roots = permitted_roots, min_size = min_size, allowed_ptms = allowed_ptms)
+      return virtual_edges, edge_labels
   return [], {}
 
 

diff --git a/glycowork/network/evolution.py b/glycowork/network/evolution.py
@@ -90,7 +90,7 @@ def distance_from_metric(df: pd.DataFrame, # DataFrame with glycans (rows) and t
   # Get all objects to calculate distance between
   value_counts = df[rank].value_counts()
   valid_ranks = value_counts.index[value_counts >= cut_off]
-  valid_networks = [net for spec, net in zip(df[rank], networks) if spec in valid_ranks]
+  valid_networks = [net for spec, net in zip(value_counts.index, networks) if spec in valid_ranks]
   # Get distance matrix
   return calculate_distance_matrix(valid_networks, dist_func, label_list = valid_ranks.tolist())