Remove HPO Recursion (#464)

* remove overcomplicated logic when searching for HPO terms related to a query term

* Bump version: 6.2.0 → 6.2.1

.bumpversion.cfg

@@ -1,5 +1,5 @@
 [bumpversion]
-current_version = 6.2.0
+current_version = 6.2.1
 commit = True
 tag = False
 

.github/workflows/index_page_builder.yaml

@@ -11,7 +11,7 @@ jobs:
     runs-on: ubuntu-latest
 
     env:
-      VERSION: 6.2.0
+      VERSION: 6.2.1
 
     steps:
 

README.md

@@ -32,15 +32,15 @@ The Dockerfile offers two builds, use the `--target` flag to specify which build
 - one for a Google cloud environment (called `talos_gcloud`).
 
 ```bash
-DOCKERBUILDKIT=1 docker build --target talos_none -t talos:6.2.0  .
+DOCKERBUILDKIT=1 docker build --target talos_none -t talos:6.2.1  .
 ```
 
 ## Workflow Demonstration with Nextflow
 
 A Nextflow pipeline is provided to demonstrate how to run Talos. This pipeline is designed to be run on a local machine, and requires that you have nextflow installed locally, and a docker daemon running. First, as described above, build the docker image:
 
 ```bash
-docker build --target talos_none -t talos:6.2.0  .
+docker build --target talos_none -t talos:6.2.1  .
 ```
 
 Then, run the pipeline:

setup.py

@@ -28,7 +28,7 @@ def read_reqs(filename: str) -> list[str]:
     name='talos',
     description='Centre for Population Genomics Variant Prioritisation',
     long_description=readme,
-    version='6.2.0',
+    version='6.2.1',
     author='Matthew Welland, CPG',
     author_email='[email protected], [email protected]',
     package_data={'talos': ['templates/*.jinja', 'example_config.toml']},

src/talos/GeneratePanelData.py

@@ -12,7 +12,7 @@
 
 import phenopackets.schema.v2 as pps2
 from google.protobuf.json_format import ParseDict
-from networkx import MultiDiGraph
+from networkx import dfs_successors
 from obonet import read_obo
 
 from talos.config import config_retrieve
@@ -89,53 +89,6 @@ def set_up_cohort_pmp(cohort: pps2.Cohort) -> tuple[PhenotypeMatchedPanels, set[
     return hpo_dict, all_hpos
 
 
-def match_hpo_terms(
-    panel_map: dict[str, set[int]],
-    hpo_tree: MultiDiGraph,
-    hpo_str: str,
-    selections: set[int] | None = None,
-) -> set[int]:
-    """
-    get panels relevant for this HPO using a recursive edge traversal
-    for live terms we recurse on all parents
-    if a term is obsolete we instead check each replacement term
-
-    relevant usage guide:
-    https://github.com/dhimmel/obonet/blob/main/examples/go-obonet.ipynb
-
-    Args:
-        panel_map (dict):
-        hpo_tree (): a graph object representing the HPO tree
-        hpo_str (str): the query HPO term
-        selections (set[int]): collected panel IDs so far
-
-    Returns:
-        set: panel IDs relating to this HPO term
-    """
-
-    if selections is None:
-        selections = set()
-
-    # identify identical match and select the panel
-    if hpo_str in panel_map:
-        selections.update(panel_map[hpo_str])
-
-    # if a node is invalid, recursively call this method for each replacement D:
-    # there are simpler ways, just none that are as fun to write
-    if not hpo_tree.has_node(hpo_str):
-        get_logger().error(f'HPO term was absent from the tree: {hpo_str}')
-        return selections
-
-    hpo_node = hpo_tree.nodes[hpo_str]
-    if hpo_node.get('is_obsolete', 'false') == 'true':
-        for hpo_term in hpo_node.get('replaced_by', []):
-            selections.update(match_hpo_terms(panel_map, hpo_tree, hpo_term, selections))
-    # search for parent(s), even if the term is obsolete
-    for hpo_term in hpo_node.get('is_a', []):
-        selections.update(match_hpo_terms(panel_map, hpo_tree, hpo_term, selections))
-    return selections
-
-
 def match_hpos_to_panels(hpo_panel_map: dict[str, set[int]], hpo_file: str, all_hpos: set[str]) -> dict[str, set[int]]:
     """
     take the HPO terms from the participant metadata, and match to panels
@@ -147,15 +100,18 @@ def match_hpos_to_panels(hpo_panel_map: dict[str, set[int]], hpo_file: str, all_
 
     Returns:
         a dictionary linking all HPO terms to a corresponding set of Panel IDs
-        a second dictionary linking all HPO terms to their plaintext names
     """
 
     hpo_graph = read_obo(hpo_file, ignore_obsolete=False)
 
-    hpo_to_panels = {}
+    hpo_to_panels = defaultdict(set)
     for hpo in all_hpos:
-        panel_ids = match_hpo_terms(panel_map=hpo_panel_map, hpo_tree=hpo_graph, hpo_str=hpo)
-        hpo_to_panels[hpo] = panel_ids
+        # identify all HPO terms back to the ontology root
+        successor_hpo_terms = set(dfs_successors(hpo_graph, hpo))
+
+        for hpo_term in successor_hpo_terms:
+            if hpo_term in hpo_panel_map:
+                hpo_to_panels[hpo].update(hpo_panel_map[hpo_term])
 
     return hpo_to_panels
 

src/talos/version.py

@@ -3,4 +3,4 @@
 """
 
 # Do not edit this file manually
-__version__ = '6.2.0'
+__version__ = '6.2.1'

test/test_metamist_hpo.py

@@ -5,7 +5,7 @@
 import networkx as nx
 from obonet import read_obo
 
-from talos.GeneratePanelData import get_panels, match_hpo_terms, match_hpos_to_panels, match_participants_to_panels
+from talos.GeneratePanelData import get_panels, match_hpos_to_panels, match_participants_to_panels
 from talos.models import ParticipantHPOPanels, PhenotypeMatchedPanels
 
 
@@ -18,17 +18,6 @@ def test_get_panels(httpx_mock, fake_panelapp_overview):
     assert panels_parsed == {'HP:1': {2}, 'HP:4': {1}, 'HP:6': {2}}
 
 
-def test_match_hpo_terms(fake_obo_path):
-    """
-    check that HP tree traversal works
-    this test is kinda limited now that the layer count is constant
-    """
-    obo_parsed = read_obo(fake_obo_path)
-    panel_map = {'HP:2': {1, 2}}
-    assert match_hpo_terms(panel_map=panel_map, hpo_tree=obo_parsed, hpo_str='HP:4') == {1, 2}
-    assert match_hpo_terms(panel_map=panel_map, hpo_tree=obo_parsed, hpo_str='HP:2') == {1, 2}
-
-
 def test_match_hpos_to_panels(fake_obo_path):
     """
     test the hpo-to-panel matching