Open cbizon opened 1 year ago
This is the query getting sent to Robokopkg that returns 0 results:
{
"message": {
"query_graph": {
"nodes": {
"n2": {
"ids": [
"DBSNP:rs7523360",
"DBSNP:rs191988426",
"DBSNP:rs3754363",
"DBSNP:rs145549455",
"DBSNP:rs1178647127",
"DBSNP:rs1674938307",
"DBSNP:rs748231565",
"DBSNP:rs11587372",
"DBSNP:rs76566517",
"DBSNP:rs142166024",
"DBSNP:rs190110651",
"DBSNP:rs369535357",
"DBSNP:rs1570953989",
"DBSNP:rs10776792",
"DBSNP:rs755485552",
"DBSNP:rs868637545",
"DBSNP:rs41312672"
],
"categories": [
"biolink:SequenceVariant"
],
"is_set": false,
"constraints": [],
"name": "Sequence Variant"
},
"n3": {
"ids": [
"MONDO:0005260"
],
"categories": [
"biolink:Disease"
],
"is_set": false,
"constraints": [],
"name": "autism"
}
},
"edges": {
"e1": {
"subject": "n2",
"object": "n3",
"knowledge_type": null,
"predicates": [
"biolink:related_to"
],
"attribute_constraints": [],
"qualifier_constraints": []
}
}
}
}
}
We're getting those 17 SequenceVariants from Service Provider
Interesting - do you get variants from ROBOKOP? Variants kind of suck because they're not in nodenorm so the service provide names won't match in RK (but the RK ones ought to)
No, get no results from the first hop:
{
"message": {
"query_graph": {
"nodes": {
"n1": {
"ids": [
"NCBIGene:7252"
],
"categories": [
"biolink:Gene",
"biolink:Protein"
],
"is_set": false,
"constraints": [],
"name": "Gene"
},
"n2": {
"ids": null,
"categories": [
"biolink:SequenceVariant"
],
"is_set": false,
"constraints": [],
"name": "Sequence Variant"
}
},
"edges": {
"e0": {
"subject": "n1",
"object": "n2",
"knowledge_type": null,
"predicates": [
"biolink:related_to"
],
"attribute_constraints": [],
"qualifier_constraints": []
}
}
}
}
}
@EvanDietzMorris why do you think we're not getting results from robokop for this query?
Sending that query straight to the plater does return results, so must be something in Strider. Note that because sequence variants don't normalize with nodenormalizer, and we always normalize sequence variants to CAID ids, nodes with DBSNP: prefixes will never return results from robokop.
Interesting. I'll look into this again.
This query connects a particular gene to autism via sequence variants.
This query runs against robokopkg and returns 3 results. Running in strider produces 0 results. I did some experiments and it seems like strider is able to get variants, so maybe there is something happening at the NN step? NN doesn't handle sequence variants, but the variant id's coming each direction from robokop should still line up.