From one RDF file containing statements about multiple chemical compounds, I would like to create one RDF file per chemical compound.
In order to do that, I started off looking for a SPARQL query that can isolate all triples linked to a certain URL, no matter how much nodes are in between.
I started off with a very interesting SPARQL query (https://stackoverflow.com/a/33290642/5433896) such that it would return (hopefully) all triples linked to a chemical compound :d1 in my dataset, but NOT about another compound :d10:
CONSTRUCT {
:d1 ?prop ?val .
?child ?childProp ?childPropVal .
?someSubj ?incomingChildProp ?child .
}
WHERE {
:d1 ?prop ?val ;
(:overrides|!:overrides)+ ?child .
?child ?childProp ?childPropVal.
?someSubj ?incomingChildProp ?child.
}
However, when I ran this on my simplified test case (python):
rdf = """<?xml version="1.0"?>
<!DOCTYPE rdf:RDF [
<!ENTITY owl "http://www.w3.org/2002/07/owl#" >
<!ENTITY owl11 "http://www.w3.org/2006/12/owl11#" >
<!ENTITY xsd "http://www.w3.org/2001/XMLSchema#" >
<!ENTITY owl11xml "http://www.w3.org/2006/12/owl11-xml#" >
<!ENTITY carcinogenesis "http://dl-learner.org/carcinogenesis#" >
<!ENTITY rdfs "http://www.w3.org/2000/01/rdf-schema#" >
<!ENTITY rdf "http://www.w3.org/1999/02/22-rdf-syntax-ns#" >
]>
<rdf:RDF xmlns="http://dl-learner.org/carcinogenesis#"
xml:base="http://dl-learner.org/carcinogenesis"
xmlns:rdfs="http://www.w3.org/2000/01/rdf-schema#"
xmlns:owl11="http://www.w3.org/2006/12/owl11#"
xmlns:carcinogenesis="http://dl-learner.org/carcinogenesis#"
xmlns:owl11xml="http://www.w3.org/2006/12/owl11-xml#"
xmlns:owl="http://www.w3.org/2002/07/owl#"
xmlns:xsd="http://www.w3.org/2001/XMLSchema#"
xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#">
<owl:Ontology rdf:about=""/>
<owl:ObjectProperty rdf:about="#hasAtom">
<rdfs:domain rdf:resource="#Compound"/>
<rdfs:range rdf:resource="#Atom"/>
</owl:ObjectProperty>
<owl:ObjectProperty rdf:about="#hasBond">
<rdfs:domain rdf:resource="#Compound"/>
<rdfs:range rdf:resource="#Bond"/>
</owl:ObjectProperty>
<owl:ObjectProperty rdf:about="#hasStructure">
<rdfs:domain rdf:resource="#Compound"/>
<rdfs:range rdf:resource="#Structure"/>
</owl:ObjectProperty>
<owl:ObjectProperty rdf:about="#inBond">
<rdfs:range rdf:resource="#Atom"/>
<rdfs:domain rdf:resource="#Bond"/>
</owl:ObjectProperty>
<owl:DatatypeProperty rdf:about="#charge">
<rdfs:domain rdf:resource="#Atom"/>
<rdfs:range rdf:resource="&xsd;double"/>
</owl:DatatypeProperty>
<owl:DatatypeProperty rdf:about="#isMutagenic">
<rdfs:domain rdf:resource="#Compound"/>
<rdfs:range rdf:resource="&xsd;boolean"/>
</owl:DatatypeProperty>
<owl:Class rdf:about="#Atom"/>
<owl:Class rdf:about="#Bond">
<owl:disjointWith rdf:resource="#Structure"/>
<owl:disjointWith rdf:resource="#Atom"/>
</owl:Class>
<owl:Class rdf:about="#Bond-7">
<rdfs:subClassOf rdf:resource="#Bond"/>
</owl:Class>
<owl:Class rdf:about="#Carbon-22">
<rdfs:subClassOf rdf:resource="#Carbon"/>
</owl:Class>
<owl:Class rdf:about="#Compound">
<owl:disjointWith rdf:resource="#Structure"/>
<owl:disjointWith rdf:resource="#Atom"/>
<owl:disjointWith rdf:resource="#Bond"/>
</owl:Class>
<owl:Class rdf:about="#Six_ring">
<rdfs:subClassOf rdf:resource="#Ring"/>
</owl:Class>
<owl:Class rdf:about="#Ring">
<rdfs:subClassOf rdf:resource="#Structure"/>
</owl:Class>
<owl:Class rdf:about="#Structure">
<owl:disjointWith rdf:resource="#Atom"/>
</owl:Class>
<Compound rdf:about="#d1">
<hasBond rdf:resource="#bond1"/>
<hasAtom rdf:resource="#d1_2"/>
<hasAtom rdf:resource="#d1_3"/>
<hasStructure rdf:resource="#six_ring-1"/>
<isMutagenic rdf:datatype="&xsd;boolean">false</isMutagenic>
</Compound>
<Bond-7 rdf:about="#bond1">
<inBond rdf:resource="#d1_3"/>
<inBond rdf:resource="#d1_2"/>
</Bond-7>
<Carbon-22 rdf:about="#d1_2">
<charge rdf:datatype="&xsd;double">-0.133</charge>
</Carbon-22>
<Carbon-22 rdf:about="#d1_3">
<charge rdf:datatype="&xsd;double">-0.0030</charge>
</Carbon-22>
<Six_ring rdf:about="#six_ring-1"/>
<Compound rdf:about="#d10">
<hasBond rdf:resource="#bond40"/>
<hasAtom rdf:resource="#d10_12"/>
<hasAtom rdf:resource="#d10_13"/>
<isMutagenic rdf:datatype="&xsd;boolean">false</isMutagenic>
<hasStructure rdf:resource="#six_ring-9"/>
</Compound>
<Bond-1 rdf:about="#bond40">
<inBond rdf:resource="#d10_12"/>
<inBond rdf:resource="#d10_13"/>
</Bond-1>
<Six_ring rdf:about="#six_ring-9"/>
<Nitrogen-32 rdf:about="#d10_12">
<charge rdf:datatype="&xsd;double">-0.313</charge>
</Nitrogen-32>
<Nitrogen-32 rdf:about="#d10_13">
<charge rdf:datatype="&xsd;double">-0.313</charge>
</Nitrogen-32>
</rdf:RDF>
"""
# Inspired by https://stackoverflow.com/a/33290642/5433896:
sparql_query = """CONSTRUCT {
:d1 ?prop ?val .
?child ?childProp ?childPropVal .
?someSubj ?incomingChildProp ?child .
}
WHERE {
:d1 ?prop ?val ;
(:overrides|!:overrides)+ ?child .
?child ?childProp ?childPropVal.
?someSubj ?incomingChildProp ?child.
}
"""
# Trying this query out:
import rdflib
import logging
logger = logging.getLogger()
logger.setLevel("INFO")
graph = rdflib.Graph()
graph.parse(data=rdf, format='xml')
result = graph.query(sparql_query)
for s, p, o in result:
print(s, p, o)
if s.endswith('#d10') or s.endswith('#bond40') or s.endswith('#six_ring-9') or s.endswith('#d10_12') or s.endswith('#d10_13'):
logging.error('This triple should not be in the results! => {0} {1} {2}.'.format(s, p, o))
I get two errors that I want to avoid:
ERROR:root:This triple should not be in the results! => http://dl-learner.org/carcinogenesis#six_ring-9 http://www.w3.org/1999/02/22-rdf-syntax-ns#type http://dl-learner.org/carcinogenesis#Six_ring.
ERROR:root:This triple should not be in the results! => http://dl-learner.org/carcinogenesis#d10 http://www.w3.org/1999/02/22-rdf-syntax-ns#type http://dl-learner.org/carcinogenesis#Compound.
I found out the answer to this question when adding extra things I had already noticed.
Since :d1 rdf:type :Compound and :d10 rdf:type :Compound, there is actually a path from :d1 to :d10 and therefore triples about :d10 also end up in my query results - which is of course not what I wanted.
Looking at the query, I thought it would make sense to add an additional constraint that ?child mustn't be :d10. And, thinking about the way the poster of https://stackoverflow.com/a/33290642/5433896 explained his query, I should also exclude that ?someSubj or ?childPropVal can be :d10:
CONSTRUCT {
:d1 ?prop ?val .
?child ?childProp ?childPropVal .
?someSubj ?incomingChildProp ?child .
}
WHERE {
:d1 (:overrides|!:overrides)+ ?child .
?child ?childProp ?childPropVal.
?someSubj ?incomingChildProp ?child.
FILTER (?child != :d10)
FILTER (?childPropVal != :d10)
FILTER (?someSubj != :d10)
}
This removed :d10 from my query results. Great!
But the error ERROR:root:This triple should not be in the results! => http://dl-learner.org/carcinogenesis#six_ring-9 http://www.w3.org/1999/02/22-rdf-syntax-ns#type http://dl-learner.org/carcinogenesis#Six_ring. remains.
I played with removal of some plausible triples causing this issue and found out that the triple { Six_ring rdf:about="#six_ring-9" } is the cause of still having #six_ring-9 in the query results. So, again, rdf:type (implied in rdf:about) is causing the problem.
Ideally, we would need describe in SPARQL that we are interested to know what the rdf:types are of objects linked to :d1 (e.g. :Compound, :Six_Ring), but NOT what other objects are also linked to those object types. That would solve BOTH problems we initially detected with the query.
So this query solves the issue:
CONSTRUCT {
:d1 ?prop ?val .
?child ?childProp ?childPropVal .
?someSubj ?incomingChildProp ?child .
}
WHERE {
:d1 (:overrides|!:overrides)+ ?child .
?child ?childProp ?childPropVal.
?someSubj ?incomingChildProp ?child.
FILTER (?incomingChildProp != rdf:type)
}