Tutorial

How to make queries on an endpoint like Pathway Commons?

BioPAX (Biological Pathway Exchange) is a RDF (Resource Description Framework)/OWL (Web Ontology Language) knowledge representation that can be stored on endpoints in two forms of graphs.

  • as Named Graph for each source of data; for example one named graph for Reactome, another for PID, etc.
  • or as an unique graph combining all data sources.

A module has been developped to provide a quick description of what users can query on an endpoint in order to forge their queries in consequence.

In a few words, three program parameters are needed to build a model:

  • The name of the queried graph: the graph URI (Uniform Resource Identifier),
  • the name of the subset of data if it exists: the provenance URI,
  • the endpoint: the the URL of the server.

Get the description of graphs on a triplestore

Example of Pathway Commons

When all databases are grouped under the same graph URI, filtering on the provenance is mandatory.

$ biopax2cadbiom model --getGraphUris --triplestore http://rdf.pathwaycommons.org/sparql/
Graph URI: http://pathwaycommons.org
        Provenance URI: http://pathwaycommons.org/pc2/biogrid
        Name: BioGRID
        Comment: Source type: PSI_MI, BioGRID Release 3.4.149 (human and the viruses), 25-May-2017
        ---
Graph URI: http://pathwaycommons.org
        Provenance URI: http://pathwaycommons.org/pc2/kegg
        Name: KEGG Pathway
        Comment: Source http://www.cogsys.cs.uni-tuebingen.de/mitarb/draeger/BioPAX.zip type: BIOPAX, KEGG 07/2011 (only human, hsa* files), converted to BioPAX by BioModels (http://www.ebi.ac.uk/biomodels-main/) team
        ---
Graph URI: http://pathwaycommons.org
        Provenance URI: http://pathwaycommons.org/pc2/pid
        Name: NCI Pathway Interaction Database: Pathway
        Comment: Source type: BIOPAX, NCI Curated Human Pathways from PID (final); 27-Jul-2015
        ---
Graph URI: http://pathwaycommons.org
        Provenance URI: http://pathwaycommons.org/pc2/panther
        Name: PANTHER Pathway
        Comment: Source ftp://ftp.pantherdb.org/pathway/3.4.1/BioPAX.tar.gz type: BIOPAX, PANTHER Pathways 3.4.1 on 04-Jul-2016 (auto-converted to human-only model)
        ---
Graph URI: http://pathwaycommons.org
        Provenance URI: http://pathwaycommons.org/pc2/reactome
        Name: Reactome
        Comment: Source http://www.reactome.org/download/current/biopax.zip type: BIOPAX, Reactome v61 (only 'Homo_sapiens.owl') 23-Jun-2017
        ---
...

Below is the command to make a Cadbiom model from Pathway Commons, based on PID:

$ biopax2cadbiom model \
--graph_uris http://pathwaycommons.org \
--provenance_uri http://pathwaycommons.org/pc2/pid \
--triplestore http://rdf.pathwaycommons.org/sparql/
  • The parameter --graph_uris provides the URI of the graphs queried (and optionally of the BioPAX ontology if it is hosted separately).
  • It is thus necessary to filter the RDF triples according to their origin with the the optional parameter: --provenance_uri. By setting it, the program will filter entities, reactions, pathways thanks to their dataSource BioPAX attribute.
  • The URL of the endpoint is specified with --triplestore.

Example of named graphs

When all databases are separated in different graphs, filtering on the provenance is NOT mandatory.

$ biopax2cadbiom model --getGraphUris --triplestore http://127.0.0.1:8890/sparql/
Graph URI: http://reactome.org/cavia
        Name: None
        Comment: http://www.reactome.org
        ---
Graph URI: http://reactome.org/cricetulus
        Name: None
        Comment: http://www.reactome.org
        ---
Graph URI: http://reactome.org/crithidia
        Name: None
        Comment: http://www.reactome.org
        ---
Graph URI: http://reactome.org/escherichia
        Name: None
        Comment: http://www.reactome.org
        ---
Graph URI: http://reactome.org/homarus
        Name: None
        Comment: http://www.reactome.org
        ---
Graph URI: http://reactome.org/mycobacterium
        Name: None
        Comment: http://www.reactome.org
        ---
Graph URI: http://reactome.org/triticum
        Name: None
        Comment: http://www.reactome.org
        ---
Graph URI: http://reactome.org/vigna
        Name: None
        Comment: http://www.reactome.org
        ---
Graph URI: http://www.irisa.fr/dyliss/data/metacyc20
        Name: None
        Comment: None
        ---
Graph URI: http://www.pathwaycommons.org/v9/pid
        Name: NCI Pathway Interaction Database: Pathway
        Comment: Source type: BIOPAX, NCI Curated Human Pathways from PID (final); 27-Jul-2015
        ---
Graph URI: http://biopax.org/lvl3
        Name: None
        Comment: None
        ---
Graph URI: http://virtualcases.org/1
        Name: None
        Comment: None
        ---
...

Below is the command to make a model from a local endpoint, hosting the PID graph:

$ biopax2cadbiom model \
--graph_uris http://biopax.org/lvl3 http://www.pathwaycommons.org/v9/pid \
--triplestore http://127.0.0.1:8890/sparql/
  • The parameter --graph_uris provides the URI of the graphs queried (and optionally of the BioPAX ontology if it is hosted separately).
  • The URL of the endpoint is specified with --triplestore.

Generated files

The conversion process will generate the model file (with a .bcx extension) in the following folder: ./output.

Unless the --no_scc_fix option is specified, two models are designed. The first is a default model corresponding to a strict conversion of the BioPAX content into a formalism based on guarded transitions. The second has specific alterations intended to optimize the operation of Cadbiom and its solver.

Strongly Connected Components

A Cadbiom model is considered as a directed graph, and is interpreted as such when it is loaded.

A directed graph is strongly connected if there is a path between all pairs of vertices in each direction. A strongly connected component (SCC) of a directed graph is a maximal strongly connected subgraph (any additonal edge breaks this property). For example, there are 2 SCCs in the following graph (we remove SCCs of 1 element to keep only “cycles”):

Two Strongly Connected Components on Cadbiom GUI

Two Strongly Connected Components on Cadbiom GUI: (i, j, k, l) and (X, Y, Z).

As it stands, there is no path from A to E without going through one of the two infinite loops (i, j, k, l) or (X, Y, Z). In order to unblock the path, we are forced to choose one node from each of the SCCs and convert it to a “start node”. The choice is systematically on the first node in the lexicographic order.

The --no_scc_fix option allows the creation of a new graph including these modifications.

Removal of SCCs on Cadbiom GUI

Removal of SCCs on Cadbiom GUI with the addition of two start nodes: i and X.

Note

Nodes selected as “start nodes” are considered as boundaries/frontier places.

Additional files

During the process 3 additional files are created in order to keep track of the changes and interpretations made to BioPAX data:

  • sort_dumped.txt: The sorted list of entities in the queried graph database. This file can also contain duplicate entities specially created for the purposes of data translation (see virtualcase11)
  • sort_grouped.txt: Groups of entities with similar characteristics.
  • sort_grouped_after_merge.txt: The final groups of entities, where we can see the definitive URI choosen for each group.