Overview

The SPARC content is as follows.

1. SPARC dataset metadata graph
   1. Datasets
      1. Publicly released datasets, including those under embargo.
   2. Protocols
      1. Hypothesis Annotations
      2. Processed Hypothesis annotations
2. SPARC Connectivity
   1. ApiNATOMY models
      1. models
         1. ard-arm-cardiac
         2. bolser-lewis
         3. bronchomotor
         4. keast-bladder
         5. sawg-distal-colon
         6. sawg-stomach
   2. Neuron Phenotype Ontology
      1. Evidence Based Types
         1. nerves.ttl
      2. NPO stubs
3. Ontology
   1. sparc-methods.ttl
   2. sparc-community-terms.ttl
   3. NIF-Ontology+

What ontologies are part of this release? The NIF-Ontology provides the foundation of the ontology used for SCKAN. The NIF-Ontology imports Uberon, Human Disease Ontology DOID, PR, and subsets of ChEBI, NCBITaxon, and NCBIGene. In addition we import the MONDO Disease Ontology, the Human Phenotype Ontology, Foundationaly Model of Anatomy, and CL.

The two releases have slightly different ontology content due to their different use cases.

1. SciGraph Everything.
2. Blazegraph Not quite everything. Only the subset that is used in the SPARC content or connectivity portions of SCKAN.

In order to create an accessible version of the Knowledge Base that can be queried we convert and enrich the SPARC content by loading it into a property graph (Neo4j) and into an triple store (Blazegraph), and by augmenting it with the NIF-Ontology which pulls in a number of community ontologies.

SCKAN = SPARC Content + NIF-Ontology + Community ontologies

Why do we have two representations?

There are two representations becuase we have found that they serve complementary use case. The triplestore is useful for executing basic competency queries over the dataset releases, but there are not existing APIs that are straight forward for devopers to consume. On the other hand, SciGraph provides a developer friendly REST API that is much easier to use in production systems.

Both of these databases are available in the docker image we provide since they are needed to run the queries. You can download the compiled versions of each database separately as well.

The SciGraph release comes as a zipped Neo4j database. The Blazegraph release comes as a journal file.

In addition to the underlying raw data, we also provide two representations of the knowledge base that can be queried directly using the SPARQL or Cypher query languages. These are available as docker images and as standalone releases.

See the README to get started querying.

Representation	Database	Language
RDF	Blazegraph	SPARQL
Property Graph	SciGraph (Neo4j)	Cypher

Neurulated groups are used to ensure that the individual segments and parts of neurons modeled in ApiNATOMY can be recognized as single cellular entities. By default ApiNATOMY treats parts of neurons individually so that it is possible to talk about the specific location of a neurite and give it an exact anatomical location.

Note however that sometimes when we talk about neurons in ApiNATOMY we implictly mean neuron populations, so a neurite or cell part is not an individual neurite of a single cell, but rather a population level representation. Cell parts here include axons, dendrites, and somas.

Population level representations can be used to generate models of individual neurons that are consistent with the population as a whole but do not differentiate between certain scenarios such as individual neurons branching vs sub-populations with distinct projection patterns.

Neurulating over the parts of populations makes it possible to recover a representation that is more familiar to those who are used to working with and thinking about whole cells.

This is useful for querying connectivity defined by neuron populations.

Neuron populations correspond sets of neurons that share defining properties the distinguish them from other similar populations. For example, there may be many populations that have their somas located in the Superior Cervical Ganglion, however they can be differentiated by considering their projection targets, both anatomically and based on their target populations.

In this knowledge base neuron populations are distinct from neurulated groups in that they are identified by the ontology representation in addition to the ApiNATOMY anatomical representation.

For the parts of the NPO that are related to SPARC, the major defining properties for the populations are the locations of their somas, axons, and dendrites. The intersection between neurite type and anatomical region is usually sufficient to uniquely identify the populations in ApiNATOMY models.

Axons and dendrites in the ApiNATOMY representation are collective unions of all the individual members of a population. This means that we do not distinguish between cases where a single neuron branches into multiple collaterals that project to different location and multiple neurons that each project to a different location and all combinations in between.

The micro-anatomy of dendrite and axonal morphology is not considered in these population level models, so any branching that is seen is representative of the macro-scale branching or differential projection patterns of whole populations.