Closed amykglen closed 1 month ago
I'll have a look
Hi @amykglen ,
After xDTD and the KPs are called, The node has two set of attributes with attribute_type_id
as biolink:synonym
. The first one seems right, but the second one seems wrong. I have confirmed that after xDTD returns results, the attributes seem correct, however after expand calls other KPs at this point, the attributes seem incorrect. As one of the synonyms is immunological adjuvant
which is part of our blocklist, this node gets removed from the KG and as a consequence of which we lose all results.
(Pdb) n1 = message.knowledge_graph.nodes["MONDO:0016262"]
(Pdb) n1.attributes[3]
{'attribute_source': None,
'attribute_type_id': 'biolink:synonym',
'attributes': None,
'description': 'Names of all nodes in this synonym set in RTX-KG2.',
'original_attribute_name': None,
'value': ['uterine leiomyosarcoma',
'Uterine leiomyosarcoma',
'leiomyosarcoma of the corpus uteri',
'uterus leiomyosarcoma',
'Leiomyosarcoma of the corpus uteri',
'Uterine Corpus Leiomyosarcoma'],
'value_type_id': 'metatype:String',
'value_url': None}
(Pdb) n1.attributes[6]
{'attribute_source': None,
'attribute_type_id': 'biolink:synonym',
'attributes': None,
'description': 'Names of all nodes in this synonym set in RTX-KG2.',
'original_attribute_name': None,
'value': ['Immunologic Adjuvants',
'Adjuvants, Immunologic',
'immunological adjuvant'],
'value_type_id': 'metatype:String',
'value_url': None}
huh! well that's odd. thanks for the debugging. tomorrow I'll look into why this node is being decorated with the wrong synonyms..
ohh, I realized what's going on - it's a bug in Plover that I'm fixing now. will deploy tonight and confirm when resolved.
alright, fixed on CI! https://arax.ci.transltr.io/?r=309955
thanks @kvnthomas98!!
https://arax.ncats.io/beta/?r=309801
this query I just ran on /beta (an 'Example 2' query for disease MONDO:0016262) finds answers, both from XDTD and from KG2, but returns no results.
the logs show plenty of answers in the KG after expand, but by the time resultify is doing its work, they're gone:
looks like it will take some debugging to figure out where they're going...
possibly related to #2395 and #2396?