This repo contains a collection of prolog utilities for querying and extracting information from NeuroLex.

Prolog has a number of benefits here:

• Fast loading of nlx triples
• More powerful programmatic/transformative/declarative access than SPARQL

This main goal of this project is to produce an OWL version of Neurolex that uses OBO Purls for classes, is suitable for reasoning, uses standard OWL axiomatic constructs (e.g. neuron-N SubClassOf soma_part_of some BrainRegion-A), removes duplication, fixes numerous nlx issues (such as duplication of concepts).

It should also be useful for querying nlx in its standard RDF form.

• http://swi-prolog.org (tested on swipl 7)
• unix make
• wget
• owltools (optional)

Type

make nlx_stage_all.rdf

This will pull the raw triples from the neurolex rdf dump (nlx_stage_all.rdf) into a local file

(This section assumes a minimal knowledge of prolog - to query the OWL scroll down to 'transforming nlx to OWL')

Type

swipl -g [nlx_util]

This will start a SWIPL session, loading the nlx_util.pl prolog code, which defines helper predicates, and also calls 'rdf_load_library(nlx)'.

The first time this is called it will take a few minutes as there is a lot of RDF to parse. This will create a directory RDF-Cache/ with a fast-cached version of nlx. This cached version will automatically be loaded on future iterations (until the RDF is re-downloaded).

Note that you will see a series of warnings. These can be ignored for now. The reason is that the neurolex property labels clash with prolog builtins. This will be fixed in future versions. Even though it looks bad, it's harmless.

The following show interactive queries that can be performed in the shell:

A unary predicate is created for every nlx class:

?- neuron(X).
X = 'http://neurolex.org/wiki/Sao2128417084' ;
X = 'http://neurolex.org/wiki/Sao383526650' ;
X = 'http://neurolex.org/wiki/Nlx_cell_20081204' ;
X = 'http://neurolex.org/wiki/Nifext_159' ;
X = 'http://neurolex.org/wiki/Nifext_152' ;
X = 'http://neurolex.org/wiki/Sao666951243' .

Other useful unary predicates:

• regional_part_of_brain/1
• parcellation_scheme_parcel/1

A binary predicate is created for every nlx property:

?- neuron(X),rdfs_label(X,L),located_in(X,S).
X = 'http://neurolex.org/wiki/Sao2128417084',
L = 'Sao2128417084',
S = 'http://neurolex.org/wiki/Category-3ANeocortex' ;
X = 'http://neurolex.org/wiki/Sao383526650',
L = 'Sao383526650',
S = 'http://neurolex.org/wiki/Category-3ACA3_stratum_pyramidale' ;
X = 'http://neurolex.org/wiki/Nlx_cell_20081204',
L = 'Nlx cell 20081204',
S = 'http://neurolex.org/wiki/Category-3ACA2' ;
X = 'http://neurolex.org/wiki/Nifext_159',
L = 'Nifext 159',
S = 'http://neurolex.org/wiki/Category-3ASubiculum' .

the capabilities here are similar to SPARQL. However, it is easier to mix in programmatic features, plus whatever rules you like, defined in prolog.

Other useful binary predicates:

• located_in/2
• species/2

The primary purpose of this codebase is to provide a "reason-able" version of Nlx in OWL. This is not straightforward due to various oddities in the RDF representation of neurolex (see Notes, below).

Type

make nlx-owl.ttl

This is the same as:

swipl -g [nlx_util],nlx2owl,halt

(no spaces after commas)

This is a work in progress. Gradually mapping more nlx predicates.

OWLTools is required for making modules. http://code.google.com/p/owltools

See the Makefile for pre-defined modules. It should be easy to add more.

Neuron module:

make nlx-neuron.owl

Nervous system part module (combo of parcels, parts of nervous system, and cell layers):

make nlx-ns-part.owl

(rough notes follow)

The neurolex triplestore has dual representations of everything.

E.g. for http://uri.neuinfo.org/nif/nifstd/sao2128417084

Sao2128417084 type 'Neuron'
    rdfs:label Sao2128417084
    nlxp:label 'Neocortex pyramidal cell'
    located_in Neocortex
        inv(sameAs) ...
    AxonMyelination 'myelinated'
    Id sao2128417084
    sameAs  Neocortex_pyramidal_cell
                 type owl:Class
                 label 'Neocortex pyramidal cell'
                 AxonMyelination 'myelinated'

We rely on the Id field to unify everything.

Need to be careful - sometimes spaces are randomly inserted, e.g.

'CHEBI: 10093'

nlx_util.pl has some hacky code for handling this.