Semantic Variation Graphs the case for RDF & SPARQL
Download as pptx, pdf0 likes579 views
The document presents the use of the Resource Description Framework (RDF) and SPARQL for representing and querying variation graphs within genomic data. It emphasizes the consistency of identifiers and the ability for diverse parties to query shared data formats. Additionally, it highlights the potential of using RDF and SPARQL in clinical applications, particularly for large genomic datasets.
![Build a “FASTA” from a VG
21
PREFIX vg:<http://example.org/vg/>
PREFIX rdf:<http://www.w3.org/1999/02/22-rdf-syntax-ns#>
SELECT ?path
(group_concat(?sequence; separator="") as ?pathSeq)
WHERE {
[] vg:path ?path;
vg:node ?node;
vg:rank ?rank.
?node rdf:value ?sequence
}
GROUP BY ?path
ORDER BY ?rank
Variation Graph as RDF](https://image.slidesharecdn.com/vg-sem-ga4gh-160518171914/85/Semantic-Variation-Graphs-the-case-for-RDF-SPARQL-21-320.jpg)
![22
PREFIX vg:<http://example.org/vg/>
PREFIX rdf:<http://www.w3.org/1999/02/22-rdf-syntax-ns#>
SELECT ?path
(group_concat(?sequence; separator="") as ?pathSeq)
WHERE {
[] vg:path ?path;
vg:node ?node;
vg:rank ?rank.
?node rdf:value ?sequence
}
GROUP BY ?path
ORDER BY ?rank
Build a “FASTA” from a VG](https://image.slidesharecdn.com/vg-sem-ga4gh-160518171914/85/Semantic-Variation-Graphs-the-case-for-RDF-SPARQL-22-320.jpg)
![23
PREFIX vg:<http://example.org/vg/>
PREFIX rdf:<http://www.w3.org/1999/02/22-rdf-syntax-ns#>
SELECT ?path
(group_concat(?sequence; separator="") as ?pathSeq)
WHERE {
[] vg:path ?path;
vg:node ?node;
vg:rank ?rank.
?node rdf:value ?sequence
}
GROUP BY ?path
ORDER BY ?rank
Build a “FASTA” from a VG](https://image.slidesharecdn.com/vg-sem-ga4gh-160518171914/85/Semantic-Variation-Graphs-the-case-for-RDF-SPARQL-23-320.jpg)
![PREFIX vg:<http://example.org/vg/>
PREFIX rdf:<http://www.w3.org/1999/02/22-rdf-syntax-ns#>
SELECT ?path
(group_concat(?sequence; separator="") as ?pathSeq)
WHERE {
[] vg:path ?path;
vg:node ?node;
vg:rank ?rank.
?node rdf:value ?sequence
}
GROUP BY ?path
ORDER BY ?rank
24
Build a “FASTA” from a VG](https://image.slidesharecdn.com/vg-sem-ga4gh-160518171914/85/Semantic-Variation-Graphs-the-case-for-RDF-SPARQL-24-320.jpg)
Semantic Variation Graphs the case for RDF & SPARQL
- 1. Jerven Bolleman Swiss-Prot Group Semantic Variation Graphs the case for RDF & SPARQL
- 4. Resource Description Framework Subjec t ObjectPredicate Resource Description Framework
- 5. Resource Description Framework Subjec t ObjectPredicate a Resource Description Framework
- 6. Resource Description Framework Subjec t ObjectPredicate a Resource Description Framework
- 7. Virtuoso Universal Server Lots of SPARQL databases Resource Description Framework ✔︎
- 10. Nodes and Edges are Resources • Resource → Identified by a URI – http://purl.uniprot.org/core/ – urn:guid:21EC2020-3AEA-4069-A2DD-08002B30309D – mailto:[email protected] – urb:isbn:978-3-16-148410-0 • Nice if public but not a requirement Resource Description Framework
- 11. Terminal edges are literals • String (xsd:string) “P53” • Date (xsd:date & xsd:dateTime) "1987-08-13"^^xsd:date • Numbers (xsd:int & xsd:decimal & …) 1 or “1”^^xsd:integer or -1.1 or “-1.1”^^xsd:decimal • Language string “Switzerland”@en “Suisse”@fr “Schweiz”@de “Svizzera”@it Resource Description Framework
- 12. Others use it too, and are cross query-able
- 13. 13 one party evolves data format everyone evolves data format Protocol Buffers Google's data interchange formatGFF
- 15. AC 4 nodes 15 ACTG T GA Variation Graph as RDF
- 16. T 4 nodes 16 1 2 4 3 AC ACTG GA base <uri of vg schema> prefix node:<uri of vg graph> node:1 a <Node> ; rdf:value “ACTG” . node:2 a <Node> ; rdf:value “AC” . node:3 a <Node> ; rdf:value “T” . node:4 a <Node> ; rdf:value “GA” Variation Graph as RDF
- 17. T 4 nodes 17 1 2 4 3 AC ACTG GA base <uri of vg schema> prefix node:<uri of vg graph> node:1 <linksForwardToForward> node:2 , node:3 . node:2 <linksForwardToForward> node:4 . node:3 <linksForwardToForward> node:4 . Variation Graph as RDF
- 18. T 4 nodes → 1 Path 18 1 2 4 3 AC ACTG GA base <uri of vg schema> prefix n:<uri of vg graph> path:1 a <Path> ; rdfs:label “Genome of patient a” ; rdfs:comment “Paths through VG make linear sequences, e.g. a reference genome or a patient assembly” Variation Graph as RDF
- 19. T 4 nodes → 1 Path → 3 Steps 19 1 2 4 3 AC ACTG GA base <uri of vg schema> prefix n:<uri of vg graph> step:1 a <Step> ; <node> node:1 ; <rank> 1 ; <path> path:1 . step:2 a <Step> ; <node> node:2 ; <rank> 2 ; <path> path:1 . Variation Graph as RDF
- 21. Build a “FASTA” from a VG 21 PREFIX vg:<http://example.org/vg/> PREFIX rdf:<http://www.w3.org/1999/02/22-rdf-syntax-ns#> SELECT ?path (group_concat(?sequence; separator="") as ?pathSeq) WHERE { [] vg:path ?path; vg:node ?node; vg:rank ?rank. ?node rdf:value ?sequence } GROUP BY ?path ORDER BY ?rank Variation Graph as RDF
- 22. 22 PREFIX vg:<http://example.org/vg/> PREFIX rdf:<http://www.w3.org/1999/02/22-rdf-syntax-ns#> SELECT ?path (group_concat(?sequence; separator="") as ?pathSeq) WHERE { [] vg:path ?path; vg:node ?node; vg:rank ?rank. ?node rdf:value ?sequence } GROUP BY ?path ORDER BY ?rank Build a “FASTA” from a VG
- 23. 23 PREFIX vg:<http://example.org/vg/> PREFIX rdf:<http://www.w3.org/1999/02/22-rdf-syntax-ns#> SELECT ?path (group_concat(?sequence; separator="") as ?pathSeq) WHERE { [] vg:path ?path; vg:node ?node; vg:rank ?rank. ?node rdf:value ?sequence } GROUP BY ?path ORDER BY ?rank Build a “FASTA” from a VG
- 24. PREFIX vg:<http://example.org/vg/> PREFIX rdf:<http://www.w3.org/1999/02/22-rdf-syntax-ns#> SELECT ?path (group_concat(?sequence; separator="") as ?pathSeq) WHERE { [] vg:path ?path; vg:node ?node; vg:rank ?rank. ?node rdf:value ?sequence } GROUP BY ?path ORDER BY ?rank 24 Build a “FASTA” from a VG
- 25. SPARQL a standard query language See VG WIKI for more examples VG 1000 Genomes → 50 GB on disk in DB VG 100,000 Genomes → ±2 TB on disk in DB Querying a Variation Graph
- 26. Summary • RDF – simple data model – consistent identifiers – anyone can say anything about anything • SPARQL – graph query language – wide scale commercial deployment – HTTP|REST in the box – in clinical use – federated queries on user demand – can be used for variation graphs
- 27. Questions? 27
Editor's Notes
- #3: I come from a strange land over the mountain where we deal with proteins ;) We even have our own 20 letter alphabet instead that 4 letter one
- #4: It’s a lot of content for 20 minutes so be sure to look at the VG wiki and ask questions on the mailing list afterwards.
- #5: Models your information in simple sentences. Key is the relationship between entities/resources/things
- #6: RDF is a W3C family of standards
- #7: You can query RDF data with SPARQL. A sql like query language for Graphs.
- #8: There is wide commercial support for SPARQL We use Virtuoso 7 a column store DB for UniProt.org See 3 Oracle logo’s its because they have 3 implementations They have different use cases. Franz has a clinical deployment with 4 trillion (10^12) edges These are some of the interesting ones with real commercial success
- #9: RDF is a conceptual model. You can serialise the same information in different ways. Each has their own usecases
- #10: The examples I show later will use Turtle, it closely aligns with the SPARQL syntax.
- #11: Key part of RDF graphs, all nodes are identified using URIs. One should be able to paste these into a browser and get some extra information (dereference). This is not an obligation.
- #12: At the edge of the graph are literals. Numbers etc…
- #13: Some are using OWL see your own ontology working group. But OWL and RDF are like two hands on one content and full belly ;)
- #14: XML, AVRO, Protobuf, Thrift => 1 party controls the xsd, schema. RDF as long as information is additive it is compatible.
- #16: I assume you have seen a VG railroad schema Here there a 4 nodes
- #17: Using the turtle syntax we describe the 4 nodes and the sequence fragments they represent. We won’t be talking about node identifiers. There is a proposal on the VG wiki.
- #18: They are simply linked in the forward to forward strand connection.
- #19: One path representing a linear sequence.
- #20: Steps say in which path they are and which rank they have. This allows us to allow arbitrary subgraphs to be serialised out while maintain a correct link to the whole graph
- #22: This is a simple query which builds a “FASTA” from the VG graph. Showing we can get the input data out again.
- #23: We want to see a ?path and its associated sequence ?pathSeq. which is concatination of all its original ?sequences.
- #24: We need to get all the sequences from the nodes
- #25: We need to group by path because there will be many. Ordering by rank is important to get the concated sequence in the correct order
- #26: See GA4GH issue for napkin math on disk usage.
- #28: While the SIB pays my salary, this work would not have happened without the NBDC/DBCLS biohackathon series. As well as the welcome by Erik Garrison and the rest of the VG team.