Ubergraph filters

[dosumis]

STATUS: DRAFT

Problem

The results of direct queries of ubergraph redundant and non-redundant graphs often give suboptimal results for biologist-facing use cases. The redundant

Case 1: Most precise object term needed.

Query of object graph => object terms that are too abstract. Query of non-redundant graph fails to => any object term in cases where redundancy stripping assumes users will be able to infer from the properties of a more general class.

Examples:

• Querying for all GO processes from some set of CL terms (e.g. via capable of)
  • Querying the redundant graph => uselessly abstract GO terms. https://api.triplydb.com/s/EYZJ7_buH

cell_ontology	GO
sensory epithelial cell	biological_process
interneuron	biological_process
motor neuron	biological_process
sensory neuron	biological_process
polymodal neuron	biological_process

• Querying the non-redundant graph gives too little, e.g.,GABAergic only links to GO BP on the most general grouping class.

https://api.triplydb.com/s/166HwhWEo

cell_ontology	GO
GABAergic neuron	gamma-aminobutyric acid secretion, neurotransmission

The redundant graph => 67 cell types

cell_ontology	GO
basket cell	gamma-aminobutyric acid secretion, neurotransmission
cerebellar Golgi cell	gamma-aminobutyric acid secretion, neurotransmission
GABAergic neuron	gamma-aminobutyric acid secretion, neurotransmission
Kolmer-Agduhr neuron	gamma-aminobutyric acid secretion, neurotransmission
rosehip neuron	gamma-aminobutyric acid secretion, neurotransmission
cerebral cortex GABAergic interneuron	gamma-aminobutyric acid secretion, neurotransmission
GABAergic interneuron	gamma-aminobutyric acid secretion, neurotransmission
...

• What we want:

cell_ontololgy	GO
fan Martinotti neuron	biological_process
fan Martinotti neuron	transmission of nerve impulse
fan Martinotti neuron	secretion by cell
fan Martinotti neuron	acid secretion
fan Martinotti neuron	gamma-aminobutyric acid secretion, neurotransmission
fan Martinotti neuron	secretion
fan Martinotti neuron	transport
fan Martinotti neuron	cellular process
fan Martinotti neuron	biological regulation
fan Martinotti neuron	regulation of neurotransmitter levels
fan Martinotti neuron	system process
fan Martinotti neuron	neurotransmitter transport
fan Martinotti neuron	neurotransmitter secretion
fan Martinotti neuron	signal release
fan Martinotti neuron	multicellular organismal process
fan Martinotti neuron	nervous system process
fan Martinotti neuron	gamma-aminobutyric acid secretion
fan Martinotti neuron	gamma-aminobutyric acid transport
fan Martinotti neuron	localization
fan Martinotti neuron	establishment of localization
fan Martinotti neuron	regulation of biological quality
fan Martinotti neuron	organic substance transport
fan Martinotti neuron	export from cell
fan Martinotti neuron	establishment of localization in cell
fan Martinotti neuron	signal release from synapse

-->

cell_ontololgy	GO
fan Martinotti neuron	transmission of nerve impulse
f
fan Martinotti neuron	gamma-aminobutyric acid secretion, neurotransmission

• Proposed Solution
  For each subject: query for all subClassOf relationships between object terms. Filter out all triples from the original query where the term has subclasses according to this second query. However, this would require many secondary queries and so would be inefficient. Is there some clever way to do this in SPARQL with subqueries?

• Add redundancy stripping functions for UberGraph queries #29

CASE2: Graph-view generation

Aim: simple redundancy stripping that does not assume users can deal with inheritance of properties down the class heirarchy.

{details and examples TBA}

[dosumis]

@balhoff - wondering if there is a way to add a third graph with less redundancy stripping than the nonredundant graph that fulfills this use case. Also would like to consider OAK as the best home for methods to take advantage of this if we can't fix with additional graphs.

[balhoff]

@dosumis what do you think about these query results? https://api.triplydb.com/s/MGJqUWP_6

[dosumis]

Not quite. e.g. the first entry here is redundant with the second in both of these cases:

Finding it hard to get my head around why this works as well as it does.

[balhoff]

@dosumis I think this one more directly encodes what you're looking for, but sometimes NOT EXISTS is expensive. It seems to work in this case: https://api.triplydb.com/s/c5DmHCzmG

and I think it gives the right results for those two terms: https://api.triplydb.com/s/SinnoFp-e

[dosumis]

That works & makes sense too. Thanks! (I think too slowly in SPARQL)

[balhoff]

Do you want to try the query approach and hold off on thinking about new graphs for the moment?

[dosumis]

I think the query approach makes the most sense & seems to scale well. We could always revisit the new graph approach in future if use-cases justify it.