-The human kidney has a complex structure and diverse interactions among its cells and cell components, both during homeostasis and in its diseased states. To better understand the kidney, it is critical to systematically classify, represent, and integrate kidney gene activities, cell types, cell states, and interstitial components. Toward this goal, we developed a Kidney Tissue Atlas Ontology (KTAO). KTAO reuses and aligns with existing ontologies such as the Cell Ontology, UBERON, and Human Phenotype Ontology. KTAO also generates new semantic axioms to logically link terms of entities in different domains. As a first study, KTAO represents over 200 known kidney gene markers and their profiles in different cell types in kidney patients. Such a representation supports kidney knowledge base generation, query, and data integration.
Kidney diseases pose a major threat to human health.
Human acute kidney injury (AKI) is a sudden and temporary
loss of kidney function. Chronic kidney disease (CKD) causes
reduced kidney function over a period of time. CKDs may
develop over many years and lead to end-stage kidney disease.
The prevalence of CKD in the general population is
approximately 14 percent [
The Kidney Precision Medicine Project (KPMP) is an NIHfunded precision medicine project aimed at finding new ways to treat human AKI and CKD. The KPMP consortium includes six recruitment sites to enroll and biopsy human subjects with CKD or AKI, five tissue interrogation sites to perform various analyses on the biopsy samples, and one central hub that is responsible for interoperable data collection, processing, visualization, and systematic analyses from the tissue- to molecular-level. A huge amount of data will be generated in KPMP; one major KPMP challenge is how to systematically integrate, store, share, and analyze this large volume of data.
In the era of big data and precision medicine, ontologies are widely used in biomedical data and metadata standardization, and robustly support data integration, sharing, and analysis. A biomedical ontology is a set of computer- and humaninterpretable terms for entities and relations among the entities in a specific biomedical domain. Hundreds of biomedical ontologies have been used in the past two decades. Together they have greatly supported the state-of-the-art biomedical research and clinical studies.
By working with the nephrology and ontology communities, we have developed a community-driven Kidney Tissue Atlas Ontology (KTAO), with the aim of systematically representing and integrating different components, cell types, and cell states of the kidney. Here we report the KTAO development strategy and how it can be used to support kidney knowledge base generation, kidney atlas data standardization and integration, and predictive data analysis to support translational kidney research.
The KTAO development follows the ontology development
principles (e.g., openness and collaboration) initiated and
promoted by the Open Biological and Biomedical Ontologies
(OBO) Foundry [
The KTAO development uses a combination of top-down
and bottom-up methods [
Existing terms from other ontologies were imported into
KTAO using Ontofox (http://ontofox.hegroup.org) [
As a first application study, we used KTAO to model and
represent known kidney gene markers collected by our KPMP
nephrology domain experts. Based on the information
available, we developed an ontology design pattern, utilized
the Ontorat tool [
The Protégé OWL editor (http://protege.stanford.edu/) was used for the KTAO visualization, manual new term generation and editing, and ontology term merging. KTAO-specific terms were generated with new identifiers using the prefix “KTAO_” followed by auto-generated seven-digit numbers. The Hermit reasoner (http://hermit-reasoner.com/) was used for consistency checking and inferencing.
KTAO is expressed using the W3C standard Web Ontology Language (OWL2) (http://www.w3.org/TR/owl-guide/). The KTAO source code is open and freely available at GitHub: https://github.com/KPMP/KTAO.
The KTAO ontology is deposited in the NCBO BioPortal website: https://bioportal.bioontology.org/ontologies/KTAO, as well as the Ontobee ontology repository website: http://www.ontobee.org/ontology/KTAO.
The Resource Description Framework (RDF) triples for the
KTAO ontology were saved in the Ontobee triple store [
Fig. 1 illustrates the upper level KTAO hierarchical
structure and selected key ontology terms of KTAO. KTAO
adopts the Basic Formal Ontology (BFO) [
(BFO) disposition (BFO) disease (DOID) function (BFO) continuant (BFO) quality (BFO) phenotype (HP)
entity (BFO) material entity (BFO) temporal region (BFO) cell (CL) anatomical
entity (UBERON)
cellular component (GO) gene (OGG) planned process (OBI) occurrent (BFO) process (BFO) life cycle (UBERONI) biological process (GO)
cellular process (GO)
KTAO imports and semantically links terms from existing ontologies (Fig. 1). For example, KTAO imports kidneyspecific cell types from the Cell Ontology (CL), anatomic entities from UBERON, and phenotypes from Human Phenotype Ontology (HPO). The terms in these different ontologies often lack linkages. One main task of the KTAO development is to link these terms together using semantic relations. Overall, KTAO aims to systematically classify, represent, and integrate different cell types in the kidney, cell states (healthy, injured, dying, recovering, undergoing adaptive/maladaptive repair, etc.), and interstitial components (collagens, proteoglycans, signaling molecules, etc.). Such an ontology development strategy also makes KTAO a basic and scalable knowledge environment for standardized KPMP data annotation, integration, and analysis.
Fig. 2 illustrates the KTAO ontology design pattern that links different types of entities in the framework of KTAO. Many of the entity types in Fig. 2 represent branches of hierarchical terms defined by specific ontologies. For example, hundreds of cells and anatomical entities are defined in the Cell Ontology (CL) and the UBERON anatomical entity ontology, respectively. While the KTAO top level design (Fig. 1) shows the hierarchical relationships among different terms, Fig. 2 shows the relations of terms across different hierarchical branches of KTAO. Therefore, the combination of Fig. 1 and Fig. 2 provides a general framework of KTAO ontological design.
As an example of KTAO development and usage, Fig. 3 illustrates how KTAO links and integrates different terms and structures from existing ontologies. First, KTAO imports kidney-related cell types from CL, anatomic entities from UBERON, human phenotypes from HPO, genes from OGG, and diseases from DOID. KTAO currently imports 259 human genes from OGG. These genes, all collected by our nephrology domain experts, are kidney disease gene markers or reference genes critical for KPMP research. Based on the reference gene panel information, we can add relation linkages (called “axioms”) showing, for example, that the WT1 gene is up-regulated in podocytes in patients with CKD, and a podocyte (also named “glomerular visceral epithelial cell”) is part of the visceral layer of the glomerular capsule (Fig. 3). Each of these entities is located in hierarchical ontological structures; for example, podocyte is under the epithelial cell branch of the Cell Ontology (CL) (Fig. 3).
Fig. 3 also demonstrates how we can provide the synonym information, i.e., ‘podocyte’ being a synonym for ‘glomerular visceral epithelial cell’. In different KPMP studies, we often find the representations of the same or similar terms using different controlled terminologies or ontologies (e.g., ICD9/10, SNOMED). We will map these different representations with our chosen ontologies using a synonym-like approach so that software programs can be developed to semantically understand these representations and the relations among them.
In addition, Fig. 3 shows the hierarchical context of different entity types. For example, ‘glomerular visceral epithelial cell’ is one type of epithelial cell; and under the same epithelial cell branch there are many other epithelial cell types, such as ‘epithelial cell of distal tubule’ and ‘epithelial cell of proximal tubule’. Such a structure allows many useful queries, such as a query of all gene markers located in various types of epithelial cells.
Note that the KTAO relation ‘susceptible to be upregulated in CKD in cell’ is generated as a shortcut relation to directly link the gene, cell, and CKD patient population, and indicates that a gene marker (e.g., WT1) is susceptible to be upregulated in a CKD patient’s cells (e.g., podocyte) (Fig. 3). The CKD in this relation represents chronic kidney disease, one of the two kidney diseases focused on in both the KTAO and KPMP. The inclusion of CKD in the relation definition simplifies the axiom representation; this is also a reason why we call it a “shortcut” relation. Similarly, other new relation terms are also generated to represent complex knowledge between different entities, such as ‘susceptible to be downregulated in CKD in cell’ and ‘susceptible to be up-regulated in AKI in cell’.
This axiom statement indicates that every WT1 is
susceptible to be up-regulated in some “glomerular visceral
epithelial cell” (i.e., podocyte) of CKD patients. The WT1 gene
encodes for the Wilm’s tumour protein (WT1), a
transcriptional factor required for podocyte development and
homeostasis [
The latest release of KTAO contains a total of 2,639 terms, including 2,357 classes, 171 object properties, and 98 annotation properties. Most terms in KTAO were imported from 31 existing ontologies. Table 1 shows a list of reused ontologies in KTAO, including BFO, CL, DO, HPO, GO, OBI, OGG, and UBERON. The usage of these ontologies is important to the full representation of the kidney atlas information in KTAO.
The full ontology statistics of KTAO can be found on Ontobee at: http://www.ontobee.org/ontostat/KTAO. As shown on the website, KTAO has many KTAO specific terms, including 13 object property terms such as the term ‘susceptible to be up-regulated in CKD in cell’ (KTAO_0000003) (Fig. 3). This term has been used to initiate 9 axioms. Other object properties such as ‘susceptible to be up-regulated in CKD in anatomic location’ (KTAO_0000009) have also been used for axiom generation. These object properties and their usages provide feasible demonstrations on how KTAO can be used to generate new axioms. More work is being conducted to add all possible axioms associated with kidney gene markers. These KTAO relation terms are critical to link together different components represented in existing ontologies.
Since KPMP has been stored in Ontobee and BioPortal, we can also query and visualize specific ontology terms and their annotations and usages in KPMP using the Ontobee or BioPortal web sites. For example, the Ontobee website http://www.ontobee.org/ontology/KTAO?iri=http://purl.obolib rary.org/obo/CL_0000653 shows the details about the cell type term ‘glomerular visceral epithelial cell’ (e.g., podocyte) (CL_0000653), including its definition, annotations, class hierarchy, and various usages (including the WT1-related semantic axiom described in the above section). NCBO BioPortal also includes the KTAO ontology information (https://bioportal.bioontology.org/ontologies/KTAO) and provides a user-friendly web query system for KTAO term browsing and searching.
The KTAO ontology is being developed with many applications in mind. First of all, KTAO is being established as a knowledge base and an environmental platform to logically and systematically classify kidney cell types, anatomic entities, phenotypes, diseases, gene markers, and biological processes, as well as the relations among these entities. Existing kidney knowledge can be accumulatively added to KTAO; for example, once we identify new kidney cell types, that information can be added to KTAO. This strategy will continuously improve KTAO and make KTAO a robust community-based framework for representing continuously generated and experimentally verified kidney knowledge in a tissue atlas.
Since the KTAO OWL format is machine-interpretable, the
generation of such a kidney atlas knowledge base will also be
easily understood by computer programs, supporting various
intelligent queries and analyses. For example, since the KTAO
can be stored in an RDF triple store, e.g., the Ontobee triple
store [
Fig. 4 demonstrates one SPARQL query over KTAO. As shown in the figure, a few lines of SPARQL query code were able to identify the gene markers known to be upregulated in podocytes of CKD patients, which has been represented in the KTAO ontology. For more practical usages, various queries can be generated with new SPARQL queries. We can also develop other software programs that embed the SPARQL query code to support additional interactive querying use cases.
In addition, KTAO is targeted to serve as a resource for KPMP data annotation, visualization, and analysis. Given the many types of clinical, pathological, and molecular KPMP experiments, it remains a huge challenge to consistently represent and annotate the large amounts of KPMP-generated data. KTAO provides a standardized terminology and controlled code system for representing kidney-related entities. If all KPMP data use the KTAO terms and codes for annotation (when needed), we can automatically integrate all the data sets from different KPMP recruitment and interrogation sites using the same semantic framework. Since KTAO logically represents the relations among different entities, KTAO also supports advanced data analysis. KTAObased standard visualization tools can also be generated to take advantage of the standard representation and logic established in KTAO.
In the emerging field of precision medicine and among various “atlas” projects, KTAO offers a novel solution that reuses and links related entities represented in existing reliable ontologies, providing a scalable and reusable knowledge atlas environment to support robust knowledge and data/metadata representation, standardization, sharing, integration, and advanced analysis.
KTAO is developed as an integrative ontology and core platform to support kidney tissue atlas application development. Instead of developing everything from scratch, KTAO reuses and integrates existing ontologies and adds community-specific terms and annotations with the same semantics and upper-level ontologies. Without the integrative KTAO platform, the terms that are extracted from existing ontologies and used in KPMP may be redundant, do not use the same semantical structure, and are difficult to be integrated and utilized. Furthermore, the renal disease community has community-specific requirements and knowledge that is more efficient and appropriate to add directly to KTAO. If newly generated KTAO terms fit well into another ontology (e.g., UBERON), we will also work with the developers of the other ontology to promptly contribute the new terms to that ontology and import back to KTAO. In this way, KTAO becomes a buffer ontology that links the KPMP domain experts and projects with existing ontologies.
We are actively collaborating with existing ontology
communities and efforts. For example, we are working with
the developers of the GUDMAP ontology, a high-resolution
ontology that describes the sub-compartments (including
histological structures and cell types) of the developing mouse
genitourinary tract [
To support the community-based ontology development, we will hold a KPMP ontology workshop this summer in Seattle with the developers of many community-based ontologies (such as HPO, UBERON, CL, and OBI) to discuss how we can better collaborate and support community-based ontology development. Such an event will become an influential platform for learning, discussion, and collaboration among experts with different backgrounds, and build up community consensuses on how to effectively develop a novel atlas ontology to support the needs in the specific kidney community and for the general precision medicine.
Diseases