Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
This article reports on the development of a novel tool for the consistent display of information linked to multiscale anatomy, such as phenotype definitions [Gkoutos et al., 2009], clinical measurements, mathematical models of disease progression, and gene expression data in an anatomical context. This graphical tool generates tiled representations of ontologies of biological structure in the form of treemaps that reflect the topology of ontology graphs. In this section, we first outline the background motivation and goals for developing such a tool, and then introduce the ApiNATOMY development effort to constrain the layout of anatomically meaningful treemaps ensuring greater visual immediacy and consistency.
Motivation and Goals
Clinical practitioners and biomedical researchers organize a considerable proportion of their data and model resources (DMRs) in terms of implicit anatomy knowledge as a basis for inferring disease mechanisms, comparing phenotype information, as well as planning experimental, investigational, and therapeutic strategies. In order to improve the manner by which such resources can be automatically integrated and meaningfully related to one another, the biomedical community has fostered the development of annotation standards for DMR semantic metadata [de Bono et al., 2011]. A central feature of this annotation standardization is the development of explicit representations of biomedical knowledge in ontological form (e.g., [Jonquet et al., 2011]). A key contribution to this explicit representation of biomedical knowledge is the communal development of reference ontologies that formally describe organ-, cellular-, and subcellular-level biological structures (i.e., multiscale anatomy). This communal effort is particularly evident in the work of the Open Biomedical Ontologies (OBO) Foundry [Smith et al., 2007], which coordinates the orthogonal development of a number ontologies that describe anatomical structures across species and scale (e.g., [Bard et al., 2005; Bard 2005; Dahdul et al., 2010; Lee and Sternberg, 2003; Rosse and Mejino, 2003; Segerdell et al., 2008]).
The annotation of DMR semantic metadata with standardized ontologies is paving the way for the use of knowledge representation inference and classification in support of automated resource management and analysis. However, a number of anatomy ontology shortcomings, involving robustness, completeness, consistency of representation, as well as application methods, are as yet an unyielding obstacle to drawing the full benefit of anatomy knowledge representation reasoning in DMR management [Travillian et al., 2011].To overcome the above limitations, additional methods must be developed that leverage knowledge represented explicitly in ontologies of anatomical structure in further support of biomedical research and the management of corresponding DMRs. One knowledge representation avenue for further study is the use of visualization techniques that enhance the effectiveness of multiscale anatomy display. To this end, the development of novel methods that utilize explicit anatomy knowledge to display ontology terms and associated semantic metadata in a visually intuitive manner would provide added value to ontology-annotated resources. For instance, the graphical integration of various types of resource metadata onto schematic two-dimensional (2D) depictions of multiscale anatomy would allow epidemiologists, clinicians, and biomedical scientists to visually review, and interact with, anatomically aggregated heterogeneous data (e.g., anatomical distribution of gene expression, signalling pathways, spread of primary and secondary tumors, etc). Such an approach would, for example, allow the rapid identification of functional relations between colocated resources that may not be immediately amenable to automation by ontology-based inferencing. In addition, the automated generation of multiscale anatomy schematics is a crucial first step for the collaborative graphical annotation of anatomy-related phenotype information by the community.
The ApiNATOMY Development Effort
In support of the above goals, we have developed a methodology that produces tiled schematics of body parts in a consistent and anatomically meaningful manner. The ApiNATOMY toolkit is a prototype JAVA application programming interface (API) generating such schematics and providing for the overlay of anatomy-related information onto the same diagram (http://apinatomy.org). This API draws upon the topology of anatomy ontologies to automatically lay out treemaps representing body parts, and makes use of semantic metadata to embed links to DMRs related to these body parts. In the present article, we focus on human body schematics.
In the sections that follow, we first introduce treemap schematics in general and discuss some of their key characteristics, limitations, and challenges. In the Ontology Treemaps section, we illustrate an approach that makes use of 2D treemap schematics to describe the human body as a three-dimensional (3D) artefact. In relation to this approach, we report on our initial results using the foundational model of anatomy (FMA) [Rosse and Mejino, 2003] to create schematics that intuitively convey the spatial and functional meaning of depicted anatomy (Depicting the Human Body Plan as a Treemap section). In Populating the Body Plan with FMA Terms section, we describe a template-based method that ensures that the latter schematics are generated in a consistent manner. In Using Templates to Ensure Consistency of Treemaps section, we demonstrate a simple graphical integration of FMA anatomical mappings from the human phenotype ontology (HPO) and the ArrayExpress Atlas of gene expression [Kapushesky et al., 2010]. This graphical integration takes the form of an HTML image map generated using the ApiNATOMY software. Finally, the implications and applications of the ApiNATOMY toolkit for the graphical navigation and interaction with heterogenous repositories of phenotype- and disease-related data are discussed.
This section provides an introduction to treemap schematics (treemaps), and their application in describing knowledge represented in ontology graphs.
Treemaps are a tiled depiction of graphs that are in the form of a simple rooted tree (see Figure 1, and refer to [Bittner and Smith, 2002] for a more formal treatment of treemaps). A simple rooted tree is an undirected graph in which any two nodes are connected by one simple path of edges such that a child node can only have one parent node. Such a tree has one root node. The graphical depiction of simple rooted trees as treegraphs and treemaps is relatively straightforward. Figure 1A introduces treegraph and treemap depictions, and illustrates the direct correspondence between them. The graph structure of a reference anatomy ontology can be drawn straightforwardly as a treemap, provided that the ontology graph is in the form of a simple rooted tree. In such an ontology graph, nodes represent ontology terms and edges represent ontology relations (the tree in Fig. 1A shows a simple example of an anatomy ontology of the eye, in which the relations are not specified). When a treemap depicts the topology of an ontology graph, a rectangle (or tile) in a treemap represents the entity that an ontology term represents (a node in a treegraph), and the placing (or “nesting”) of a tile inside another represents a parent–child relation between concepts represented by ontology terms (an edge in a treegraph). In treemaps, metadata mappings can also be symbolically overlaid onto the appropriate ontology term tile (e.g., experimental data generated from different regions of the eye). For example, DMRs linked by metadata to ontology terms are represented in triangular form in Figure 1A.
However, the use of treemaps to depict anatomy knowledge has two significant limitations. The first pertains to knowledge loss due to treemap creation. The graph structure of most anatomy ontologies (such as the FMA) does not consist of simple rooted trees, but of directed acyclic graphs (DAGs). In DAGs, child nodes are allowed to have more than one parent node. This implies that an ontology DAG has first to be converted into a simple rooted tree before a treemap can be generated without difficulty. This conversion step is likely to lead to significant loss of knowledge from the reference ontology.
The second limitation pertains to the inconsistent depiction of treemaps drawn using different portions of the same large ontology graph. For instance, the FMA has tens of thousands of term nodes. In practice, therefore, the typical visualization use case is for treemap depiction to be carried out on an arbitrary subset of FMA terms. A graph created from this subset of terms is called a derivative subgraph (Fig. 1B). If the resulting treemap of such a subset of nodes is to preserve the original knowledge associated with the FMA reference ontology, then the derivative subgraph bearing these nodes must faithfully preserve the topology of the FMA reference graph from which it is extracted. The automated extraction and layout of two derivative subgraphs from the same reference graph is shown in Figure 1B. In this diagram, the two subgraphs have three terms in common (i.e., nodes #1, #26, and #31, colored in blue) and their respective layout is achieved through the application of classic treemap layout algorithms (e.g., [Bruls et al., 2000]). The figure shows that the generated spatial arrangement of the three communal term tiles is considerably different between the two treemaps. Such discrepancy suggests that a tile representing the same body part may be placed inconsistently and erratically across a treemap, subject to its location within the particular derivative subgraph in which it is embedded. A treemap produced in this manner is not easily compared with other treemaps derived from the same reference ontology. The use of inconsistent treemap layout in a graphical user interface environment results in user disorientation and frustration, particularly when the underlying derivative subgraph is updated during real-time interaction. If treemaps are to be usefully applied to the real-time visualization of reference ontologies of biological structure and associated metadata, the relative position of tiles representing particular ontology terms, therefore, must be constrained to ensure consistency of depiction over all possible derivative subgraphs from the same reference graph. A technical solution to this problem is discussed in Populating the Body Plan with FMA Terms section.
Depicting the Human Body Plan as a Treemap
Treemaps are classically drawn in two dimensions, whereas, as a geometric entity, the human body is in three dimensions. Therefore, ease of anatomical treemap interpretation is dependent on the process by which 3D anatomy is projected onto the 2D treemap. This section explores one avenue for effective projection that imbues anatomy treemap schematics with spatial and functional meaning in a manner that is normally achievable only with 3D depictions. Although the solution proposed here illustrates one of many ways in which anatomy treemaps may be organized for the effective display of associated biomedical knowledge, the suggested overall technical approach ( Depicting the Human Body Plan as a Treemap section) and implementation (Populating the Body Plan with FMA Terms section), discussed in later sections, are generally applicable.
A solution for the intuitive conversion from 3D to 2D anatomy is to reduce the geometric complexity inherent to the bilaterally symmetrical coelomate body plan (Figure 2A), to which humans conform, into an idealized radially symmetric one (Fig. 3B). In doing so, the gut lumen of this simplified body plan provides the central longitudinal axis of rotation around which cocentric strata of body parts are radially disposed.
Figure 2B shows one such idealized stratum that was arbitrarily apportioned into six contiguous cylinders (numbered 19–24) arranged cephalocaudally. Given that the simplified human body plan is radially symmetrical, any arbitrary plane that is parallel to, and passes through, the central axis of rotation of this stratum creates a longitudinal section. This plane, therefore, contains a perpendicular tile to each corresponding cylinder (six tiles in Fig. 2B, 24 in Fig. 3B). This set of contiguous 2D tiles gives rise to a planar asymmetric unit of the 3D body plan. An asymmetric unit can recapitulate the entire symmetric 3D body plan through a rotational operation on this unit about the central longitudinal axis.
Figure 3B proposes an example of a whole-body asymmetric unit (wbAU) that organizes an idealized body plan over 24 cocentric cylinders, each representing a distinct region of the body. The relative position of the corresponding wbAU tiles within the asymmetric unit carries additional functional meaning as follows.
1.All wbAU tiles are composed of nonoverlapping body parts, and each tile represents a body region that carries out a well-established physiological function. The incremental numbering of the tiles requires the wbAU to represent the longitudinal section of the body in the supine position while looking at the body from its left-hand side (Fig. 3A).
2.Body regions represented by surface wbAU tiles (i.e., tiles in the periphery of the wbAU that share a boundary with either two or three other tiles) bear epithelium and, as such, are involved in exchange processes with the environment.
3.Surface tiles 19–24 bear solely the epithelium of the skin, while tiles 1–7, 12, 13, and 18 bear mucosal and nondermal epithelium.
4.Body regions represented by core wbAU tiles (i.e., tiles that share a boundary with four other tiles) represent endothelial and neural systems. Tiles 8–11 bear cardiovascular and lymphatic vessels (apart from microcirculation) and include the spleen, endocrine glands, and bone marrow. Tiles 14–17 bear all neural tissue (apart from peripheral motor end plate and sensory receptor tissues) and include the cerebrospinal fluid and associated meninges. Tile 18 is not in this core category, but represents the main cranial sensory structures, and includes the eyeballs (bearing conjunctival epithelium), as well as olfactory and vestibulocochlear apparatuses (bearing upper respiratory epithelium).
5.In most cases, when two wbAU tiles share a border, the body regions in those two tiles are engaged in well-established processes of exchange between them. This is particularly true for any core tile with any other tile, as well as surface tiles 1–6, 12, and 18.
6.Tiles 19–24 bear all skin, bone, and connecting skeletal musculature, and their longitudinal arrangement models the typical route of musculoskeletal stress transfer while standing.
7.The relative position of tiles 1, 7, and 13 reflect the canonical topology of the anal, vulval, and urethral orifices in the female perineum, respectively.
The above approach sets the stage for the application of the wbAU to organize the distribution of anatomical terms over a 2D treemap schematic that intuitively conveys spatial and functional meaning customarily linked to a 3D cylindrical body plan. The next section illustrates our initial results in mapping relevant portions of the FMA to the wbAU, while ‘Using Templates to Ensure Consistency of Treemapsxs’ with FMA Terms section discusses the computational methodology adopted in generating such treemaps automatically and consistently.
Populating the Body Plan with FMA Terms
The aim of the work described in this section is the redistribution of anatomical knowledge within the FMA according to the allocation heuristics embodied by the 24-tile wbAU approach outlined in Ontology Treemaps section. Specifically, entire portions of the FMA are extracted and assigned uniquely to the relevant wbAU tile. The initial focus of our efforts is the subset of about 27,000 FMA terms that represent parts of the body (the “Body” concept is represented by the FMA term with ID 256135). This large subset of the FMA will be called the “map set,” and each term in the map set graph can be reached from the “Body” root node via a unique path consisting of any combination of the following edge relations: regional part, constitutional part, and direct subclasses. For the purpose of this work, the FMA DAG bearing the map set was converted into a simple rooted tree, with the above ordered list of relations indicating the priority with which relations were preserved during the conversion process.
The next step involved curation that applied the above heuristic rules to uniquely assign portions from the FMA map set graph to individual wbAU tiles. Each of these 24 tiles could be “seeded” with any number of FMA terms (Fig. 4A). In practice, this curation step involved building an anatomy reference graph from scratch to create a modified FMA tree. The FMA “Body” term is the root node of this reference graph (as shown in Fig. 4A), and it is linked to 24 wbAU tile nodes via regional part relations. The same type of relation then linked “seed” FMA terms to the wbAU tile nodes. The assignment of a seed term to a tile node led to the elimination of any prior relations the seed term had with its erstwhile parent terms in the original FMA ontology. One such example is illustrated in Figure 4B, where the seeding of tile node #24 with node 58 “FMA:Face” would also, by extension over the partonomy relation, have assigned the node 1 “FMA:Mouth” to the same tile. As this term and its components belong, functionally, to tile #6, the manual seeding of node 1 to tile #6 automatically uncouples “FMA:Mouth” from “FMA:Face” (this uncoupling is depicted as a red “X” sign in Fig. 4B). As a layout example, Figure 4C shows the correspondence between the treegraph and treemap associated with tile #6, together with an illustrative overlay of metadata associated with the anatomy terms.
In total, 27,432 FMA terms were uniquely linked to 24 wbAU tiles, as outlined in Supp. Table S1.
Using Templates to Ensure Consistency of Treemaps
In this section, the key design considerations and ApiNATOMY toolkit implementation for the generation of consistent FMA treemaps are discussed.
The previous section showed how thousands of anatomy terms were uniquely allocated to individual top-level tiles of the wbAU, as a requisite step for whole-body treemap generation. In practice, it is unlikely that the average user will require the depiction of all 27,432 FMA term tiles all the time. In fact, by drawing only those term tiles that are required, an effective use of space by the treemap is ensured.
As a simple use case scenario, as illustrated in Figure 5, consider the extraction of a target set of FMA terms from the curated reference graph (described in Depicting the Human Body Plan as a Treemap section) to create a derivative subgraph. One such derivative subgraph may initially consist of the following target set of four FMA terms, namely, “transverse colon,” “parotid gland,” “tibia,” and “right pinna” (strictly speaking, the resulting derivative subgraph would also contain the FMA term “Body” as root by default, as shown in derivative subgraph #1 in Fig. 5A). The curation effort described in Depicting the Human Body Plan as a Treemap section resulted in the four FMA terms in this target set being respectively assigned to wbAU tiles #1 (large intestine), #6 (mouth and throat), #19 (lower limb), and #24 (head) (see correspondence between Fig. 3B and C for wbAU tile names and numbers). If the functional meaning associated with the wbAU arrangement (discussed in Ontology Treemaps Section) is to be reflected and preserved in the resulting treemap, then the layout of this derivative subgraph has to be constrained in accordance to the spatial knowledge embodied by the wbAU. Such constraint would ensure that:
1.no space in the treemap is needlessly allocated to wbAU tiles that are unpopulated (20 wbAU tiles, in the example in Fig. 5A, 18 in Fig. 5B), and
2.specifically, the “transverse colon” term tile is placed in the top-left corner of the treemap, the “parotid gland” tile is in the top-right corner, “tibia” tile is in the bottom-left corner, and “right pinna” tile bottom-right corner, as specified in the wbAU. An example of such a treemap is shown in subgraph #1 in Fig. 5A.
More generally, if the spatial knowledge in the wbAU is formally encoded, then it becomes possible to computationally constrain the spatial layout of term tiles for any derivative subgraph originating from the same curated reference graph. Consider the outcome of adding two new FMA terms, “uterus” and “trigone of urinary bladder,” to the previous target set of four terms. The new derivative subgraph (subgraph #2 in Fig. 5B) would result in a six-tile treemap being generated. If wbAU layout constraints were imposed, the four corners of the new treemap would still be occupied by the same four term tiles as before. In addition, the left-side border of the new treemap would be tessellated as follows in the downward direction: “transverse colon,” “uterus,” “trigone of urinary bladder,” and “tibia.” Layout constraints, therefore, support the stable and visually consistent transition between derivative subgraphs from the same reference graph.
The constrained layout of the derivative subgraphs shown in Figure 5 is the result of the ApiNATOMY toolkit making use of a graph-based template associated with the FMA “Body” term node. This template formally encodes the spatial knowledge embodied in the wbAU, as depicted in Figure 3C, such that the wbAU template consists of a mesh of 24 nodes—one for each wbAU tile (see Fig. 6). The wbAU template represents the desired topological features of the wbAU in graph form. For instance, the ApiNATOMY software prepares for the six-tile layout shown in Figure 5B by first laying out the treemap for the whole 24-node mesh template. In a subsequent step, those six template tiles bearing FMA terms from derivative subgraph #2 are expanded in order to take over the space left over by the 18 unpopulated (i.e., empty) template tiles (the direction of expansion is indicated using blue arrows in Fig. 5B).
In practice, the application of treemap layout constraints need not be limited solely to the general organization of the 24 functional regions of the body represented by the wbAU. Given a curated reference graph of more than 27,000 FMA terms, a template can be usefully associated with any parent term that has two child terms or more (11,245 nodes in all in this particular graph). The multilevel association of templates with parent terms in the same reference treegraph is illustrated in Figure 6. In effect, the wbAU template is but one of a number of templates that can be associated with parent terms in the reference graph. For instance, the local layout of the child terms of the “stomach” term may also be additionally constrained to ensure that the relative 2D arrangement of any subset of child terms within the “stomach” tile bears the meaning encoded by the “stomach” template in consistent manner. The application of multiple templates for the consistent layout of treemaps is illustrated in Figure 7, where the result of generating two treemaps from very similar FMA derivative subgraphs is shown. In this figure, the layout of the two treemaps was, in fact, constrained by two templates, not just one. As previously established, the first template imposes the topological constraints of the wbAU arrangement. The second template is associated with the FMA parent term “stomach.” This template constrains the layout of all nine FMA child terms in the reference graph that are regional and constitutional parts to the “stomach” term. Five of these nine child terms are laid out under the constraint of this template in Figure 7A. In this figure, the child nodes and tiles of the “stomach” term under the additional local influence of the second template are colored in light grey on the darker background of the gastrointestinal regions.
The key role templates play in ensuring consistency of treemap depiction between different derivative subgraphs of the same reference graph is illustrated by comparing Figure 7A and B. Each of these two figures shows a distinct derivative subgraph, where the subgraph in Figure 7B is the result of removing 10 nodes from the subgraph in Figure 7A. In effect, the following nodes were removed from the derivative subgraph in Figure 7B: three vascular wbAU tile nodes and their FMA term content (nine nodes in all), plus a term from the stomach wbAU tile, namely, “body of stomach.” In both Figure 7A and B, the blue arrows indicate how gastrointestinal and neural tiles are directed by the wbAU template to expand vertically to take up the space previously occupied by the vascular tiles. Similarly, the yellow arrows indicate how the template associated with the “stomach” term constrains the layout of the remaining four tiles to preserve a similar 2D spatial relationship these same four tiles had in Figure 7A.
It is also worth noting here that without the template-based constraint of tile layout, classic treemap algorithms offer limited guarantees of consistent layout between similar derivative subgraphs. Figure 7A and B offers a case in point with regards to the inconsistent layout solution for the “arch of aorta”(numbered “1” in both figures) and “common carotid artery” (numbered “2”) child nodes to the “vascular cephalic” wbAU tile. As the two FMA terms are not under the influence of a local template, the slight change in the aspect ratio of the “vascular cephalic” tile, between the ratio in Figure 7A and in B, prompted the unconstrained algorithm to alter the way the corresponding internal areas for the two FMA terms were apportioned. Given this type of unstable and inconsistent behavior, graphical user interfaces that only provide unconstrained treemap generation are not likely to be of much practical use, especially when dealing with large derivative subgraphs that are constantly being updated.
Integrating Phenotype Data into Treemaps
By way of demonstration of the ApiNATOMY methodology in the context of phenotype-related resource management, a wbAU-constrained “phenotype map” consisting of relevant FMA terms was generated. The list of phenotype-related FMA terms was derived from the set of phenotypic class definitions created by the HPO effort [Robinson and Mundlos, 2010]. The resulting image (illustrated and color-coded Supp. Figure S1) was created in the form of an HTML image map to allow basic user interaction with the information being displayed. This image map is available as Supporting Information (and at www.apinatomy.org/material), and allows the user to “mouseover” the map to retrieve the following three types of related data, namely:
1.the text label and identifier of the FMA term involved in the definition of a particular HPO class;
2.the text label and identifier of HPO classes associated with a specific FMA term; and
3.gene expression data that is annotated with the FMA term represented by the corresponding tile, as retrieved from ArrayExpress Atlas [Kapushesky et al., 2010].
The image map was generated using an internal development version of the ApiNATOMY toolkit, and the following was noted:
1.The HPO class source file, human-phenotype-ontology_xp.obo, was downloaded from the compbio.charite.de site on September 14, 2011;
2.Nine hundred and eight FMA terms were found to be mapped to HPO classes in the above source file;
3.Of the 908 FMA terms referred to by HPO classes, 586 were part of the curated reference graph discussed in Depicting the Human Body Plan as a Treemap Section.
In practice, the HPO ApiNATOMY map consists of a derivative subgraph of 586 FMA terms that were distributed over 22 tiles of the wbAU template. The HPO treemap shows that, of the 586 FMA terms depicted at http://apinatomy.org and Supp. Figure S1, a preponderance of tiles was musculoskeletal in nature.
In preparing this diagram, we did not take into account the precise meaning of the relationship between anatomy and phenotype terms. One of our future research aims is to develop automated reasoning mechanisms that elicit and depict anatomy-to-phenotype mappings more accurately.
A significant proportion of biomedical resources carries information that cross references to anatomical structures, such as reference to:
1.cellular, tissue, or organ structures involved in some process both as participants in, and location of, such a process;
2.the anatomical location from which some measurement was taken; and
3.complex clinical and disease concepts that are, in part, defined in terms of multiscale anatomy participants or location.
The annotation of semantic metadata that links DMRs to formal biomedical knowledge representation is paving the way to automate the management of resources through machine-readable representations of anatomy. To this end, the automated generation of visual maps of multiscale anatomy can play a key role in DMR metadata management, as well as contribute directly to biomedical data analysis and interpretation of results.
The ApiNATOMY methodology described in this article provides anatomically meaningful and consistent visualization of biological structure. Specifically, ApiNATOMY draws upon multiscale anatomical knowledge explicitly represented in ontologies to generate anatomy treemaps in an automatic manner. In so doing, ApiNATOMY maps can contribute to the visual integration of DMR metadata and complex clinical concepts.
Treemaps are well suited to represent partonomy relations (i.e., relations that describe the composition of the parent in terms of its children). Above all, this type of relation is particularly well represented in ontologies of biological structure compared to other types of spatial relation. However, there is more to an effective treemap than the mere conveying of partonomic knowledge through tile nesting. If properly constrained, the relative position of child tiles that belong to the same parent can serve as additional means to imbue the diagram with further spatial meaning.
In ApiNATOMY, templates associated with parent terms provide such means of constraint. Furthermore, if additional nonpartonomic spatial knowledge is readily available from an ontology of a biological structure, then it is feasible for parent-associated templates to be automatically imported into ApiNATOMY from that ontology. However, as sourcing nonpartonomic spatial information is less likely, the independent curation or automated generation of ApiNATOMY templates also provides a feasible way for the biomedical community to enhance ontology treemap layout in a communal and customized manner. It is also envisaged that different biomedical communities may develop discipline-specific templates. For instance, the view of pelvic functional anatomy for an orthopaedic surgeon may be distinct from that of an obstetrician, such that the two groups may want to develop their own standardized layout of the pelvic parts.
The ApiNATOMY toolkit development plan does encompass the production of tools (such as template editors) to simplify and support the utilization of the software by the community. Indeed, technical discussions about its algorithm, planned ApiNATOMY software releases, as well as its application in depicting structures at different levels of scale, will be further developed in separate publications and documented on the ApiNATOMY website.
In this article, we have discussed the rationale and application of one whole-body template that describes the longitudinal asymmetry of an idealized human body—a radially symmetrical homunculus. While the development of such a template, as well as its seed annotation by FMA terms, is still in progress, the spatial and functional knowledge reflected by this template is of interest for the following two key reasons:
1.the template apportions the body over 24 major tiles (Fig. 3C), where each tile represents regionally colocated tissues that, in most cases, are involved in convective flow or stress exchange processes with tissues in neighboring tiles. Such a map provides a potential route to project resources generated for systems biology research onto a physiology-based view of the body;
2.the geometric derivation of the template (Fig. 3B) allows for a smooth graphical transition between 2D and 3D projections of anatomy-related resources. The option of resorting to 3D cylindrical extrusions to allay the space limitation of an overcrowded planar layout is feasible. This extrusion is feasible because the constraints associated with the 2D template can be smoothly carried into the third dimension to ensure that the cylindrical distribution of information is still meaningful. Indeed, the allocation of radial coordinates to such projections may carry meaning in its own right.
ApiNATOMY makes novel use of knowledge explicitly represented in ontologies of biological structure to present and integrate (1) resources and (2) related clinical concepts with greater visual immediacy. The “phenotype map” generated in this work illustrates a starting point for the incremental, community-driven sharing of meaningfully annotated phenotype-related information with the aim of creating a more detailed map of inherited human disease. The central role community-supported ontologies play in the generation of such maps serves to emphasize the importance of (1) data interoperability standards and (2) application ontologies such as the HPO that build complex concepts in terms of communal anatomy ontologies. In the longer term, community-driven requirements that inform the ApiNATOMY development plan are focused on supporting clinical and scientific graphical user interfaces and dashboards for biomedical resource management and data analytics.
Disclosure Statement: The authors declare no conflict of interest.