The 'Insect Brain Database' (IBdb) can be found on the internet at insectbraindb.org and is freely available to everyone. It can be used with most modern web browsers with active Javascript (tested with Google Chrome, Safari, Firefox) on computers running any operating system, without the need for any additional plugins. A user account can be registered free of charge and is required for users who wish to download data and to contribute content.

The IBdb is divided into three main hierarchical layers: Species, brain structures, and neuron types. Each level is additionally linked to 'experiments', which is the fourth major organizational layer of the database (Figure 1A). At each level, a database entry is represented by profile pages, on which all relevant information is collected (Figure 1C). These profile pages are the core of the database. They can be reached directly by several search functions, as well as by entry lists. As profile pages are embedded in a hierarchical framework of species, brain regions, and cell types, they become automatically associated with metadata. For instance, an entry for a pontine neuron of the fan-shaped body (central body upper division) in the monarch butterfly would become linked to the brain regions 'fan-shaped body/central body upper division' and its parent region 'central complex', as well as to 'monarch butterfly' as species. The neuron can then be found, for example, by querying the brain regions associated with it, or by exploring the species of interest.

Figure 1 with 2 supplements see all

Download asset Open asset

Species entries contain representative data of an insect species, with the aim of defining that species and the overall layout of its brain. Similarly, entries for brain regions and cell types contain representative examples of data that illustrate the respective entity and provide all information to unambiguously define it, essentially providing type-specimen. While the definition of species and brain regions are straight forward in the context of this database, neuron types can theoretically be defined in many different ways, based on connectivity, function, developmental origin, morphology or neurochemical identity. Given that the database is predominantly organized according to anatomical principles, we have defined a cell type as a collection of individual neurons that are morphologically indistinguishable at the level of light microscopy. In many cases, a cell type comprises only a single individual per brain hemisphere, but in cases where multiple identical neurons are present, a cell type is defined as the first level of similarity beyond the individual neuron. Higher levels of neuron categories group neurons according to anatomical similarities (Figure 1—figure supplement 2).

Contrary to the first three levels, experiment entries contain specific data from individual, defined experiments (Figure 1B). As research data can be obtained at the levels of species, brain regions and cell types, experiment entries exist for all three levels and are accessible via the respective profile pages (Figure 1A). This distinction between representative and concrete data is important, as for example only one entry for a certain columnar neuron type of the central complex exists in the database, yet, if that single cell type was subject to 30 intracellular recordings, the profile page of that cell type would list 30 experiments, each containing a unique individual neuron with its associated physiology. While an anatomical type specimen can easily be defined as the most complete example of any particular neuron’s morphology, the decision of what content to depict as representative functional data is less straight forward. The functional information on the profile page should reflect the range of experimental results present in the experiment entries of any particular cell type. As those results potentially diverge, for example due to different experimental paradigms used in different research groups, the function section can contain as many entries as needed to capture the available information, without need for consensus. With new experiment entries added, this section can evolve over time.

In the context of this evolving content, it is key to ensure that entries which are cited in published work remain findable in the exact form they existed when they were cited (Ito, 2010). Therefore each entry receives a persistent identifier. This identifier (a 'handle') links to a version of the entry that was frozen at the time it was created (and cited). Once information is added or removed from that entry, a new handle must be generated, providing a new access link for future citations. This system is applied to multiple levels of the database (experiments, neurons, and species) and ensures that all information in the database, as well as the interrelations between entries, are truly persistent.

As locating specific datasets is one of the core functions of a database, we have developed a novel, more intuitive way to find specific neuron data. A graphical representation of the insect brain, resembling the overall anatomical outline of all brain regions and highlighting the regions containing neuron data (Figure 2A), makes it easy to search for neurons within single species and across species. This graphical interface is generated directly on the database website for each species and is adapted from a generic insect brain, that is a shared ground plan. This generic brain is the least detailed fall-back option for any cross-species search and was generated based on the insect brain nomenclature developed by Insect Brain Name Working Group et al., 2014. It resembles the shared anatomical hierarchy of all brain regions in insect brains. Within this hierarchy, the entire brain is divided into 13 super-regions, which consist of individual neuropils. The latter can be further divided into sub-regions. While all super-regions exist in all species, differences become more pronounced at lower levels of the hierarchy. The generic brain therefore largely contains super-regions as well as several highly conserved neuropils (Figure 2—figure supplement 1, Figure 2—figure supplement 2). As these categories are simply tags of brain region entries used to organize the database, the search interface does not differentiate between them, simplifying the user experience (Figure 2A).

Figure 2 with 2 supplements see all

Download asset Open asset

If more than one species is subject to a query, a schematic brain is generated that displays the commonly shared features of the species involved. For both single and multi-species search, when selecting a specific brain region, all neurons in the IBdb that connect to this region become visualized by a dynamically drawn wiring diagram (Figure 2B). Filters can be applied to narrow down search results according to neuron polarity, functional class, etc. Individual neurons in the wiring diagram can be selected to reveal the neuron's profile page. Here, all available information for this cell type is displayed, including links to deposited experiment entries (Figure 2C). The schematic display of search results can visualize any neuron in the database, only requiring that a neuron is annotated with respect to the brain regions it innervates. A list of search results is additionally made available after each search and offers the possibility to also display experiment entries associated with the found neurons.

For single species queries, two more modes for visualizing search results are available: the semi-schematic view and the 3D view. The semi-schematic view mode emphasizes the natural brain organization on the level of brain regions, while also serving as interface for launching search queries. It comprises a full series of automatically generated sections through a segmented 3D brain of a species. Each brain region present in that species' 3D brain is shown as an interactive cross section that can be used to query neural connections of that region (Figure 2D). If a region is selected, all connected brain areas are highlighted and neurons resulting from this query can be visualized by changing to either the schematic view, the 3D view, or a list view. The advantage of the anatomically correct layout of this interface is that a brain region can be queried for a neuron, even if its name is not known to the researcher. This is particularly useful for regions with uncommon names that have only recently been introduced to the insect brain naming scheme (e.g. crepine, superior clamp, etc., see Insect Brain Name Working Group et al., 2014). Launching a search for neurons in regions with unfamiliar names is made much easier when the search interface resembles the information a researcher has obtained from, for example, confocal images or physical brain slices. The semi-schematic mode of the database search function fulfills this demand and bridges the schematic wiring diagram view and the full 3D view.

The 3D view visualizes search results in an anatomically correct way and shows queried neurons in the context of a species' reference brain (given that this information was added) (Figure 2E). It displays interactive surface models of that brain together with neuron skeletons obtained from the neuron-type's profile page.

In the graphical search interface the search parameters are limited to anatomical information defining the neuron's location in the brain (i.e. likely input and output areas). In contrast, an additional text-based search function (’Expert Search’) allows users to query all information deposited on a neuron's profile page. Individual search parameters can be logically combined to generate arbitrarily complex searches. The results are displayed as a list of neurons, which can be sent to populate a schematic wiring diagram view by a single click. Thus, this tool effectively combines complex search with the advantages of an intuitive display of results.

A different means of locating data on neurons and experiments is achieved via a publication based search function. Each neuron and experiment that is associated with a publication becomes automatically part of a dataset linked to this publication. By definition, experiments are only part of one publication, while neuron entries can be referenced by many publications. In either case, users can locate all data that has contributed to a specific piece of scientific literature.

Finally, whereas the emphasis of the database search lies on locating cell types, information on brain regions can also be found using identical interfaces. The schematic search option allows users to reveal brain region profile pages by selecting schematic neuropil representations. The same information can also be obtained by selecting brain regions in the semi-schematic neuropil search interface.

To maximize the usefulness of the database, we have implemented an integrated 3D viewer to deliver platform independent, high-quality data visualization without any additional software demands. Neuropil visibility can be independently switched on and off for each brain region, transparency can be freely adjusted, and colors of neurons can be changed (Figure 3A). Neurons can be shown either with diameter information or as simple backbones. The built-in screenshot function enables the user to capture any scene displayed in the 3D viewer and produces a high-resolution, publication-ready image with transparent background (Figure 3B,C).

Figure 3

Download asset Open asset

The IBdb allows users to not only locate neuronal morphologies quickly, but also to combine arbitrary neurons from any single species into a common visualization. To achieve this, we have generated a neuron clipboard, in which individual neurons from search results can be stored temporarily (Figure 3D). Any subset of cells in the clipboard can be sent to the 3D viewer, as long as all neurons belong to the same species, that is can be displayed using the same reference brain. The desired configuration of neurons and neuropils can be generated using the interactive tools of the viewer and the screenshot function can be used to create a high-resolution image to be used for illustration purposes (e.g. reviews, conference talks, teaching).

Additionally, we have embedded a function to directly compare up to four neurons side by side on screen. Any neuron located in the neuron clipboard can be chosen to be included in this comparison. The comparison uses the 3D view, the profile image, or the confocal stack located on the respective neurons' profile pages. This function is ideally suited to compare homologous neurons from across species to quickly assess differences and shared features of these cells. The four-window 3D viewer retains all functions of the normal full screen 3D viewer and thus also allows the capture of high resolution screen shots of each of the neurons being compared (Figure 3E).

The data in the database are suited for many applications, including more sophisticated ones. To provide direct access to all levels of the data in the IBdb we have created an API interface, specifying how to automatically draw data from the database via web-browser based apps. Applications produced by third parties that use this function can be embedded directly on the IBdb website, once they are approved by the site administrators. Applications envisioned are, for example, quantitative comparisons of both single neuron morphologies and neuropils between species, direct online multi-compartment modeling of neurons deposited in the database, or virtual reality interfaces that allow exploration of anatomical data in a 3D virtual reality environment. Over time, we hope that our unified platform will stimulate the insect neuroscience community to generate a collection of online tools to analyze and explore neuroanatomical and physiological data deposited in the IBdb, thereby allowing straight-forward meta-analysis of all raw data deposited in the database. As an offline tool, the Natverse package in R already offers the possibility to explore IBdb data (Bates et al., 2020).

All data on the internet is public. This also applies to any data publicly available in the IBdb. Driven by the requirement to obtain data persistency and implemented by the use of handles, no data can be removed from the database once it is public (and thus citable). For all data, the contributors retain ownership and hold the copyright to their data. The publishing is performed explicitly by the owners, not by the database administrators, and the license attached to each dataset is a Creative Commons Attribution Non Commercial 4.0 International (CC BY NC 4.0). Thus, when data in the IBdb are downloaded for reuse, the original work that underlies these data has to be credited together with the IBdb as the source. When data are used to generate images with the help of the IBdb, these images are licensed via Creative Commons Attribution 4.0 International (CC BY 4.0), that is they can be used in any publication as long as the original data owners and the IBdb are credited.

Data can be contributed by registered users at all levels of the database, i.e. species, brain regions, neurons, and experiments. The process is similar on all levels but requires more expert knowledge the higher in the hierarchy the data reside. In the following sections we will briefly illustrate the main principles of how to contribute data (for full instructions see the Online User Guide).

Species

To submit a new species, a profile page has to be created, which then has to be populated with data. While photographs, bibliography, and text descriptions of the species are desirable and strongly encouraged, the most important next step is to generate a schematic brain (Figure 4A). This will ensure that brain structure entries are created in the database, a prerequisite for neuropil-based search and 3D brain region identity. To generate a schematic brain, we have created the 'Brain Builder' tool on the IBdb website, which provides templates based on either the generic insect brain, or related species already deposited in the database. The user can simply copy an existing brain, associate it with the new species and modify it to match any unique features of the new species. Note that the ventral nerve cord (VNC), even though not part of the brain, was included as a single region as well, allowing to deposit information about VNC neurons as well and opening the possibility to expand the database beyond the actual brain in the future.

Figure 4

Download asset Open asset

Once the profile page and the schematic brain is created, a 3D brain can be uploaded to illustrate the brain organization of the new species and to serve as reference brain for neuron display (Figure 4A). For all uploaded data, the database distinguishes between source files and display files. Source files contain the 3D reconstruction in a format that the researcher would like to make available for others in the field. The second set of files, the display files, are required for automatic online display and must constitute the surface models of each neuropil (.obj-format). Each brain region model is tagged with a unique neuropil identity, so that the schematic brain regions and the 3D surface models will be linked to the identical brain-structure entry. Finally, following the same principles as for the 3D brain, an image stack can be uploaded to the profile page as well (Figure 4A). This can be any representative dataset that illustrates the layout of the species' brain (e.g. confocal stack, µCT image series, serial sections with any other technique).

Neurons

Contribution of neurons follows a similar procedure as the contribution of new species (Figure 5). The user creates a profile page for the new cell type that subsequently has to be populated with information. To make a neuron findable in the database, its arborization regions in the brain have to be defined (Figure 5C). Within a graphical user interface, these regions are chosen from the brain structures available in the schematic brain of the respective species. One arborization entry has to be created for each branching domain of the neuron, leaving no part of the neuron un-annotated. To enable the automatic generation of a wiring diagram view of the new neuron for displaying schematic search results, an outline of its branching structure has to be generated in an embedded tool called the neuron-path assistant (Figure 5D). This branch tree defines which neuropils are innervated in which order and where main branch points are located.

Figure 5

Download asset Open asset

The remaining procedure for neuron contribution is largely identical to species contribution and follows the dual approach towards source data and display data for 3D reconstruction and image stacks (Figure 5B). All other information, that is images, bibliography, keywords, representative functional data, transmitter content, and textual descriptions, can be added to the profile page at any time prior to publication. To ensure common minimal standards and to avoid rudimentary datasets, several data fields have to contain valid information before a request for publication can be made. These include an image of the neuron that allows to clearly identify the cell type, a complete morphology section (including soma location, a description of the neuron’s morphology, and arborization regions), as well as a link to a publication (or preprint) that contains data on the cell type depicted in the new entry. If no publication is available, a note in that section will state that the neuron is not part of any publication.

Importantly, the IBdb can be used to house data that have been obtained by classical methods, for example camera lucida drawings of Golgi impregnated neurons. While no 3D information is available in those cases, drawings can be uploaded as images after which annotation of the neuron's morphology (arborization regions) is performed as described. These neurons will therefore become findable in the schematic search interface and will be added to the publicly available pool of neuronal data. To allow these datasets to exist in the database, 3D reconstructions and image stacks are not mandatory content. Similarly, 3D data that is not registered to a common reference frame can be deposited in the database, with or without individual brain region reconstructions, but a shared display in the context of a reference brain will not be available for these datasets.

The database is managed via a group of voluntary curators, a scientific administrator, and a technical administrator. Importantly, no single person curates all data in the IBdb, but each species is managed by a specialized curator, who is an expert for that species. This distributed curation system ensures that no single person is responsible for too many datasets, and that no curator has to evaluate data outside their area of expertise. To additionally reduce the workload for species with many entries, more than one curator can be assigned to any given species. The scientific administrator (the lead author of this publication) oversees the curators, while technical administration is carried out by the technical administrator (last author of this publication). The technical administrator is the only person who has potential access to all data in the database.

The responsibility of the scientific administrator is to approve new species and to train and support the curators for individual species. This training is carried out during a training period during which actions of the curator have to be approved by the scientific administrator before they take effect. Once a new curator is sufficiently trained to carry out all tasks independently, the scientific administrator grants full curator privileges. The responsibility of each curator is to approve new neuron-type datasets and to re-evaluate major updates of these (data re-approval). This process entails checking for formal errors in the submitted data, ensuring that new data do not accidentally duplicate already existing data, and that the quality of the data meets acceptable standards. To ensure swift correction of any issues, we have implemented a private communication channel between curator and data owner. This was realized through a commenting function that enables the curator to post comments on a neuron page, which are only visible to the data owner. The owner can then directly respond to the comments and any issues raised can be resolved.

While our distributed approach to curation has many advantages, it creates challenges towards ensuring that curation of data across all species and levels is accurate, consistent and complete. We have therefore designed multiple tools and features to facilitate effective and consistent data curation. Besides formalized training and approval for each new curator, we have provided checklists for both data contributors and curators (found in the online help menu) that must be followed to ensure that database entries fulfill predefined standards. Yearly meetings among all curators will ensure that these standards are known and can evolve over time. If mandatory data is not provided by a data contributor, request for approval of the dataset is automatically blocked, preventing rudimentary datasets to enter the public section of the database. Finally, if an entry becomes obsolete, for example when new research data provides conclusive evidence that cell types listed as separate entries belong to the same type, existing entries can be archived. This process preserves the persistent handle of the entry, but removes it from public lists and search results in the IBdb. This way, references to these datasets remain resolvable, while, at the same time, obsolete data cannot clutter the database. In the long run, this function provides the means to keep the database clean without violating the principle of data persistence.

The definition of what constitutes minimal standards for a neuron or species entry reflects a scientific consensus among curators. Rather than imposing these standards onto the field, the IBdb generates the means to develop a common set of rules by providing the platform for continued discussions among curators as well as to reinforce an evolving consensus.

To enable all users to provide feedback and to discuss topics relevant to other database users, we have added a discussion forum directly to the IBdb website. This forum is intended as a means for reporting potential bugs, suggesting new IBdb features, or for discussing scientific content (methods for data processing or acquisition, requests for literature, staining protocols etc.).

Each database entry has to be explicitly published by the contributor. In the process, it is approved by either the database administrator (species), or the species' curator (neurons). While this procedure was initially intended only as a quality control measure to prevent incomplete or inaccurate data from compromising the database, we have developed it into a unique feature: the IBdb private mode. Before a dataset is made public, it is invisible to all other users, curators and the scientific database administrator. The dataset can thus be updated and even deleted. This creates the potential of using the IBdb to deposit data while they are being collected or prepared for publication in a research paper, that is, for data management (Figure 6A). The API can be used to programmatically integrate the IBdb into any individual data management workflow; a community driven MATLAB implementation for automatic deposition of cell types and experiments is available on GitHub (https://github.com/zifredder/IBdb-matlab).

Figure 6

Download asset Open asset

To facilitate the use of the IBdb as a data management tool, we have enabled three operational modes of the database site: private, public, and mixed. Any user logged in can thus choose to either access only (own) private data, only public data, or both. The first mode turns the IBdb into a data management site for ongoing research, the second mode is the default mode for viewing publicly available data, and the third mode allows the user to compare their own unpublished data with public data.

As efficient data management requires researchers from the same laboratory, as well as collaborators, to have access to relevant unpublished data, each user can grant access to their own private datasets (Figure 7). To this end, a user can create a user group and invite other database users to join. Datasets can then be added and made visible or editable to all members of the group. These data can either comprise individual entries at all levels of the database, or collections of entries defined by common features.

Figure 7 with 1 supplement see all

Download asset Open asset

Finally, users are often reluctant to make datasets available to the public before they are included in a research paper, yet, these data should be available to editors and anonymous reviewers. We have therefore created the possibility for 'pre-publishing' database entries (Figure 6A). This function allows a user to assign a persistent handle to a dataset (e.g. a neuron profile page) without approval by the curator and without making the dataset findable through search or lists in the IBdb. Editors and reviewers of the research paper (anyone in possession of the handle) then have direct access to the linked pages. Once the manuscript is accepted for publication, the user can submit the respective datasets for publication, obtain curator approval, and thus make them findable in the IBdb public mode.

To avoid having to separately provide numerous independent handles when sharing data, individual entries can be grouped into datasets. These receive a unique link that grants access to the entire collection. Only public or pre-public data can be included in datasets.

Public entries of the IBdb are maintained in a dual way; the persistent version is locked and cannot be changed, whereas a second, current version remains visible to the owner and to all members of user groups with appropriate access rights (Figure 6B). This current version is fully private and can be freely edited or expanded. Importantly, no data that is already part of a public version can be deleted. Rather, when for instance a confocal image stack should be replaced by a better one, the old stack can be archived, so that it will not be visible in new versions of the dataset, but will remain present in the database for display of earlier versions. Once all required updates of a dataset have been made, the edited version can be re-published and will be assigned a new version of the persistent identifier after re-approval by the assigned curator. This new version is now also locked and any further edits will again have to be done in the current version of the dataset. This ensures that all data that have been assigned a persistent identifier will remain valid and accessible, while at the same time allowing each entry to evolve. The described strategy of publishing and re-publishing and the associated duality of persistent and current versions are implemented at the levels of species, neurons and experiments.

The IBdb does intentionally not include Drosophila melanogaster as a species. This is because huge efforts have been spent developing highly efficient resources for this widely used model system and, as a result, the Virtual Fly Brain (VFB) resource has been created (Osumi-Sutherland et al., 2014). It bundles data from several older Drosophila databases (e.g. FlyCircuit) to the most recent connectomics datasets (Scheffer et al., 2020). Serving as the main repository for anatomical data from the Drosophila brain it has become the main site to locate GAL4 driver lines, single-cell morphologies, and synaptic connectivity data. It contains tens of thousands of datasets and is designed to specifically meet the needs of the Drosophila research community. By being less specialized, the IBdb has a wider scope. We are hosting many species and include both functional and anatomical data. We also do not require neuronal anatomies to be registered to a reference brain, if this is not possible for some reason. This opens the IBdb up to more diverse data, but as a result cannot provide most of the specialized services that VFB can deliver (e.g. automatic bridging registrations of 3D data between different reference brains). Cross-linking the two databases to effectively enable comparing neurons from the insect species deposited in the IBdb with Drosophila, nevertheless, was highly desirable. Importantly, both databases have converged on highly similar hierarchical frameworks. As the brain nomenclature used by the IBdb and VFB is identical, neuropil identities were mapped across both databases. In cases were homology between corresponding neuropils is unclear, regions map to the next higher order brain region (e.g. Monarch butterfly dorsal bulb maps to the entire Drosophila bulb). For some neuropils, there are no known counterparts in the fly (e.g. the posterior optic tubercle).

We use the APIs of both databases to allow direct communication between the IBdb and VFB. In principle, when searching the IBdb for neurons, an API mediated query can be automatically sent to VFB and search results are displayed as a list of single neuron entries. Each item on the list contains information obtained from VFB and is directly linked to the corresponding entry at VFB. This feature is available for the graphic, brain region based search. This includes complex multi-neuropil searches, aimed at identifying neurons connecting several brain regions. Effectively, this enables a user to launch a query in the IBdb and directly view corresponding neuron data in VFB, including data from recent connectomics efforts (Scheffer et al., 2020). This seamless interoperability makes maximal use of both complementary resources, without duplicating functionality.

Previous and current online databases hosting insect neuroscience data have suffered from several shortcomings: Most severely for old databases, a lack of maintenance often quickly led to outdated file formats, rendering the deposited information no longer compatible for viewing with current web browsers (anticipated by Ito, 2010). Second, while some databases originally allowed interactive viewing of the data, no possibility for contribution of one’s own data existed and data download was limited to very few files, for example Brandt et al., 2005, Kurylas et al., 2008. Usability of larger databases was generally impaired by a layout that often required expert knowledge to be able to launch meaningful database queries or to understand search results (e.g. FlyCircuit, Invertebrate Brain Platform [now: Comparative Neuroscience Platform]). This not only applies to old databases, but the restriction to individual species and often highly complex interfaces limit the potential user base to specialists even in cutting edge databases such as VFB or visualization tools such as FruitFlyBrainObservatory. Finally, the limitation to purely anatomical data, including in the major current cross-species database NeuroMorpho.org, does not account for one of the key advantages of insect neuroscience: the high level of tractable structure-function relations.

The IBdb addresses each of these issues. Firstly, we have developed the database software to be independent of the operating system and type of web browser used, as well as to not rely on any third party plugins. Additionally, we implemented the database as a classic, relational database without experimental data structures (e.g. intelligent, adaptive search), aiming at maximal robustness. Having created a conceptionally simple software that uses standard web-technology with standard file-formats makes continued compatibility and technical maintenance comparably simple.

Second, we have invested substantial effort in making the IBdb intuitive to use and visually attractive to provide a positive user experience. The latter factor should not be underestimated in its importance. One of the problems encountered in previous databases were user interfaces that were difficult to use, creating an immediate negative experience when attempting to use a site for the first time and therefore reducing the motivation to interact with it. As for commercial software, we aspired to generate a user interface that is largely self-explanatory, provides immediate visual feedback when a user action was successful, and which clearly shows what actions can be performed. Several years of beta-testing by a multitude of users have streamlined the site to a point at which interacting with the IBdb is both straight forward and fun.

As the success of the database depends on many users sharing their data, we aimed at making contributing data as intuitive and as easy as searching and visualizing data. We have thus simplified the data contribution process to a point where only very limited anatomical knowledge is needed, aiming at enabling physiologists without deep anatomical training to submit data as well.

Finally, to our knowledge the IBdb is the first database in the field of invertebrate neuroscience that combines functional data with anatomical data, and at multiple levels ranging from entire brains to single neurons. This ability, together with the possibility to deposit not only representative data but concrete sets of experiments, provides an opportunity for anatomists to directly interpret their findings in a functional context, as well as allowing physiologists to tether their findings to a coherent anatomical framework that automatically generates context for any functional data. The ease of use of the IBdb, combined with housing functional and anatomical data, has the potential to facilitate interactions between expert anatomists and physiologists and thereby strengthen structure-function analysis across the diversity of insect brains.

The landscape in the insect neuroscience research community has changed dramatically, and most relevant funding bodies are in the process of implementing open data mandates or have already done so (e.g. European Research Council, National Institutes of Health, Wellcome Trust, etc). Thus, the initial driving force for data deposition is much larger compared to that surrounding earlier database attempts. Yet, why should researchers use the IBdb for meeting these new mandates rather than other available databases? Different from open databases, for example Figshare.org, the IBdb is dedicated to insect neuroscience and thus provides all tools required to manage, annotate, and cross-link the specific data formats generated in this field. However, it is not rooted within a single laboratory, nor a single species, thus providing a framework for data from a broad research community. The unique possibility for cross-species comparison and the combination of anatomical and functional data additionally broadens the relevance of the IBdb.

Visualization of anatomical research data is often difficult and particularly complex when involving data from other research groups. We have implemented a range of tools enabling the visualization of data in fast, flexible, and effortless ways. This saves considerable time compared to other available software tools, in particular for complex 3D neuron data (e.g. Amira). The data contributed are also immediately incorporated into the framework of existing data. Outside the IBdb these data are distributed across many publications. Comparison of one’s own data to any published data would entail contacting authors, obtaining files in unpredictable formats and finding ways to compare them to one’s own work within the software a research group is currently using. The IBdb solves these issues and delivers such comparisons within seconds. Crucially, these advantages are already present immediately after data upload, prior to publication. Via the private mode of the IBdb, individual neurons can already be compared to their counterparts in other species while datasets are being obtained, enabling the user to generate visualizations suitable for conference contributions and publication figures. This possibility is to our knowledge unique to the IBdb and provides a major motivation for data contribution.

Importantly, the dual function of the IBdb as data repository and data management tool eliminates the need to reformat and prepare datasets for publication, a process that is required when submitting datasets to websites dedicated to only data deposition. The IBdb therefore provides a streamlined and integrated experience from data acquisition to publication, aimed a minimizing researcher workload. This is additionally relevant as formulating a data management plan has also become mandatory for projects funded by most funding organizations. By using the IBdb, users in insect neuroscience not only have access to a dedicated data repository, but at the same time have a tool at hand for efficient handling of ongoing research data, making data management plans easy to design and effective. This is particularly the case when using the API for accessing the data, as up- and download can be automated and custom designed interfaces can be programmed according to the needs of any particular research group.

Independent from these aspects, the API of the IBdb offers users the opportunity to expose their published data directly to approved third party applications. The bundled availability of data from many research groups generates a possibility of data reuse with a much broader scope than any individual solutions, providing an increased motivation for developers to design workflows that incorporate deposited data. High-quality data deposited in the IBdb therefore additionally increases the visibility of the data owners.

While the data currently deposited in the IBdb represents only a starting point to illustrate the functions and possibilities of our database concept and user interface, the neurons and species present in the IBdb already highlight new biological insights as well as concrete paths towards such insights.

Most obvious insights can be gained from data deposited in the IBdb that is not published or publishable elsewhere. For example, isolated neural morphologies from the Monarch butterfly have been published solely in the IBdb and demonstrated a previously unknown connection between the gall of the lateral complex and the mushroom body lobes (dp-GA-MB(L) neuron, NIN-0000383), thus directly linking the output of the central complex with the mushroom body for the first time in insects.

The possibility for direct comparison of neuropil and neuron data from many insect species has the potential to identify discrepancies in neuropil definitions across species and pinpoint solutions for revised homologies. For example, the main output neurons of the central complex (PFL or CPU1 neurons) are most likely homologous across insects and target the dorsal LAL in all species in which this region has been described (e.g. butterflies, moths, flies, beetles), except in locusts, in which they terminate in the ventral LAL. This discrepancy suggests that the definition of subregions of the LAL and their borders to surrounding brain regions (e.g. crepine) might have to be re-evaluated.

Along similar lines, the IBdb has generally the potential to expose discrepancies in functional data. Work carried out in a research group interested in sensory processing might define the function of a particular cell type very differently than a group focused on motor control, or one focused on neuromodulatory functions, simply because different experimental paradigms are aimed at different aspects of neural function and interpretations of results might be biased toward the underlying hypotheses. Such diverging views of neuron function can silently coexist in the literature for long times, but if bundled in side-by-side function entries on the same neuron profile page in the IBdb they become highly obvious. Such exposure will facilitate scientific discussion and the emergence of unified ideas of neural function.

Finally, in any field it is important to consistently define technical terms, agree on key concepts, and have consistent standards regarding what is acceptable data to validate conclusions. During the development phase of the IBdb it became clear that the isolated work in many insect model species has led to a wide range of cases in which similar terms have different meanings in different species and where views on what is acceptable proof for, for example, neural function varies. The definition of cell type is one such example, where researchers working in different species and brain regions attached substantially different meaning to the term. With the unifying approach of the IBdb we had to therefore identify common ground and define cell type in a way that was acceptable to each contributing party. Similar issues for other aspects of insect neuroscience research will likely be exposed by the increased levels of communication enabled by the IBdb. While resolving these issues will often take time and effort (and will involve researchers from across the field), the main function of the IBdb in this context is not to impose a strict set of rules, but to provide the framework in which inconsistencies can be exposed and resolved, eventually facilitating biological insight.

The IBdb is designed for long-term accumulation of data by many contributing research groups across the field of insect neuroscience. To successfully provide this service, the IBdb has to address several challenges of sustainability. These challenges are threefold: First, ensuring technical and financial maintenance of the database infrastructure; second, guaranteed, continued scientific oversight and expert curation, and third, lasting scientific relevance. We have therefore taken steps to address each of these points.

In the light of fast changing web technology, maintenance of high technical standards requires continuous effort and active updating of the database code, which in turn requires financial resources. The first issue is covered by an ongoing agreement with the web-developers that built and maintained the database software over the period of the last six years. Financially, voluntary contributions from research groups that initiated the database have paid for its creation. As the maintenance costs are a small fraction of the development costs, it will be easily possible to run the database within the framework of the existing service agreement for at least the next 5 years without any changes required. However, when the data volume increases substantially, the static costs of housing the data will increase accordingly. While keeping all public, persistent data available free of charge is mandatory (given that the IBdb functions as a public data repository), maintaining the IBdb as a free data management tool, that is allowing unlimited private data for each user, will likely become unsustainable over time. If this becomes a problem, free space in the private section of the database will have to be restricted. All space required beyond a certain limit will have to be rented to directly offset the costs for maintaining and administering these data. At the same time, research groups involved in creating the IBdb use this platform as their primary tool for management of research data and public data repository. The obligation to formulate data management plans and strategies for public deposition of research data, combined with the lack of equally suitable alternative platforms, will ensure that third party funding dedicated to maintaining the database will continuously be available via research funding of the founders of the IBdb.

To anticipate the slowly growing costs of housing the database due to increasing data volume, we aim at eventually relocating the data from the currently used commercial Amazon cloud platform to an academic server that is provided at minimal costs or free of charge. To this end we have ensured that the IBdb does not depend on any core functions of the Amazon cloud storage service, enabling to move the database to a new location with comparably moderate effort.

The second issue of continued quality control and expert curation is addressed at several levels. Firstly, at the level of species, the scientific administrator has the main responsibility to oversee the curation of the species entries by each species’ dedicated group of curators. Given the limited number of species that can realistically be included in the IBdb over the coming years (our estimate is a maximum of several hundred), the associated workload for general oversight is limited and manageable by a single person. As species curators actively perform research on the species they are responsible for, there is a substantial self interest to maintain high standards to advertise their work and thus facilitate their research.

At the level of neuron entries, the workload is higher due to larger data volumes. Accordingly, the main responsibility for oversight is more distributed and lies with the species’ curators. Once datasets exist in the database, updating is generally optional and will only in rare cases be necessary. In those cases, the strongest incentive for keeping data up to date lies with the data owners and research group leaders, who also possess the highest expertise for these data. Overall, expertise is mostly required to approve new datasets, a process requiring substantial overview over the data available for the relevant species. With an increasing number of species, the number of curators will also have to increase. This makes curator recruitment and training key requirements both for quality assurance and for enforcing consistent curation strategies. For this purpose, regular meetings of all curators will be held and approval of new curators will be carried out only after an extended training period. To attract new curators in the future, the IBdb administrators will actively approach researchers in the field of insect neuroanatomy. Finally, to identify existing quality issues with deposited data, establish routine workflows, to support curators and contributors at all levels, and to recruit new users, a dedicated, full time database curator position is being created, initially funded by members of the IBdb consortium for a minimum of one year.

Third, continued scientific relevance of the database will have to be ensured to attract users and contributors long term. Most crucially, after introducing the database to the field, the available deposited information has to grow beyond a critical point, at which the IBdb becomes the natural choice for depositing insect neuroscience data. We believe that this point has already been reached, as the number of deposited data sets is growing extensively (Figure 7—figure supplement 1) and users without affiliation to the founding consortium have begun to deposit data. Nevertheless, attracting more data will remain a key mechanism to increase the attractiveness and acceptance, and, consequently, the relevance of the database in the field. As one of the key advantages of our concept is the possibility to deposit all anatomical and functional data from insect neuroscience, irrespective of data format, we have begun to approach authors of published work to enable them to make old data available to the growing IBdb user base.

While highly useful for classroom teaching of structure-function relations in neural systems, the IBdb has proven to be invaluable to introduce new members of a research team to the basic layout of brains, neurons, and neural circuits in a particular species. Using the database serves as an easy (and fun) access point to available information on a research species, including key publications, and therefore saves significant effort when writing review papers, PhD thesis introductions, background sections for travel grants, etc. While this is true for established researchers as well, it is especially true for younger scientists and students who are new to the field or are at the beginning of their careers.

Finally, the IBdb provides the possibility for anyone to access original research data in intuitive and attractive ways (Figure 7). This provides opportunities to design teaching assignments for neuroscience students to carry out meta-analyses. With access to the data in the IBdb via the application programming interface (API), we have provided the possibility for third parties to develop dedicated teaching tools that provide streamlined methods to use the data for specific classroom exercises. Beyond researchers and students, journalists, interested members of the public, or members of funding bodies can also view and explore neuroscience data. Ideally, this will contribute to a more transparent understanding of what the output of science is and could spark increased interest in insect neuroscience.

The framework we have generated with the IBdb is not limited to housing insect brain data. Without major modifications it would be equally suited for hosting data from other animal taxa. While the intuitive, schematic search engine would not be useful for comparing species that do not share a common basic brain outline (i.e. a relevant 'generic brain'), the text-based expanded search could allow the construction of queries across multiple groups, for example searches according to functional terms. We are currently conceptualizing the expansion of the database toward including spiders and envision that crustaceans and other arthropods would be logical next groups.

The IBdb therefore provides not only a tool for the insect neuroscience community to facilitate data management, data visualization, transparency of results and effective teaching, but it can easily be expanded toward related fields. Additionally, it might also serve as a blueprint for how to set up similar databases in unrelated research areas. In principle, the strategies used in the IBdb are applicable to any scientific field that can be linked to a hierarchical framework.

All web infrastructure is hosted by Amazon web services on servers located in Frankfurt, Germany. Data is stored using a PostgreSQL relational database hosted by the Amazon Relational Database Service and files are stored using the Amazon S3 object storage service. The servers hosting the website and the local HANDLE system are running in Amazon EC2 containers, which runs Linux. Resources communicate using Amazon Virtual Private Cloud.

The database structure and interaction is managed by a python based Django application. User authentication, permissions, and data security are also managed within the Django application. A NGINX web server hosts static content and serves as a reverse proxy for dynamic content served by a uWSGI application server hosting the project's Django application. Asynchronous tasks are implemented using the Celery distributed task queue and RabbitMQ message broker.

Data is externally accessible via a web API delivering content in the JSON format to the front-end web application. The web API was implemented with Django and the Django REST framework.

Long-term data persistency is provided to allow users to reference information or profile pages on the site in scientific publications and other external media in a static state, while continuing to allow data to be updated as more information is acquired. When a request is made by a user for a persistent copy of a dataset to be created, a copy of the data related to the current state of the dataset is serialized and parsed into JSON. A persistent unique identifier is then assigned (HANDLE). The JSON data, HANDLE and additional metadata is recorded in a separate table and can no longer be modified. All files associated with the persistent dataset are marked as locked in the database and can no longer be modified by the user. The recorded state of that dataset can be accessed and viewed on the site using the url associated with the assigned HANDLE. The original data copied to create the persistent dataset can be modified without effecting the persistent dataset. Additional files may also be added, but will not be reflected in earlier persistent records.

The front-end of the database is primarily implemented using the Angular web framework in Typescript, HTML and SASS (CSS extension language). The Typescript, HTML and SASS are compiled and bundled with the Angular CLI using WebPack to create the distributed application files targeting ECMAScript 2015 capable browsers. Graphical consistency is targeted for browsers using Webkit and Gecko-based layout engines adhering to web standards.

The web based three-dimensional viewer was implemented using Typescript, WebGL and the Three.js three-dimensional graphics framework. The two-dimensional schematic view, brain designer and path designer was implemented using Typescript, the Canvas API and Paper.js vector graphics scripting framework.

User to server communication is protected by the Hyper Text Transfer Protocol Secure (HTTPS). User authentication is managed by the Django authentication system. Access to data is restricted by object based permissions limited to authorized users through the web interface or API arbitrated by the Django server.

File downloads of protected content stored on Amazon S3 are accessed using time limited urls, assigned to an authorized user at the time of a download request by the Django server. Files are directly downloaded from the S3 storage to the user using the temporary URL. The process is seamless to the end user who needs only be logged into the website and click the link associate with the intended file. Downloads are logged and accessible to the owner of the data being accessed. Files can be uploaded to the S3 object storage only by authorized users with a time limited URL provided by the server. The upload is logged and associated with the contributing user. Publicly available thumbnail images and other reduced quality images are stored separately in a publicly accessible (read-only) S3 bucket.

Content explicitly designated as public is accessible through the graphical user interface or via the API to any authorized visitor. Private content is only accessible to authorized users given permission to access the data.

The PostgreSQL database is protected by a firewall allowing access only via the Amazon Virtual Private Cloud and is not open to direct access via the web.

All names for brain areas are in line with previous research. The brain regions of the generic insect brain follow the new insect brain nomenclature introduced for Drosophila by Insect Brain Name Working Group et al., 2014. Accordingly, we have established three hierarchical levels of brain regions: super-regions, neuropils, and sub-regions. Super-regions are stereotypical and can be expected to comprise the ground pattern of the brain in all insects (although some might be reduced in certain species). The only exception to the Insect Brain Name Working Group et al., 2014 scheme is the anterior optic tubercle, which we have raised to the level of super-region, given its prominence and distinct nature in most insect species. Sub-regions are often specific to individual species and therefore, if such regions were defined, we used the names given to them within the relevant species. We did not unify for example names of the mushroom body calyx divisions across species, as this would firstly imply homology where there might not be any, and second, novel naming schemes will have to be developed by the community and not be imposed by a data repository. Anticipating that changes to brain names can and will happen in the future, all names, as well as the level of a region within a hierarchy, can be modified.

Within some neuropils, regular, repeating elements can be found, usually defined as columns and layers. We have implemented such a system in the central complex, that is without having to define an array of sub-regions, several strata and orthogonal slices (following the new brain nomenclature) can be generated. The default number of slices in the generic brain is 16, assuming that this number is the ancestral state of this region.

Neuron names follow the conventions within each species, as there is no common naming scheme for insect brain neurons yet in place. However, we provide the possibility to define several alternative names for each cell type to allow the parallel use of names. This is possible as the identity of a neuron is linked to the persistent ID, and not to the neuron's name. Given that we house neurons from multiple species, we add a prefix to the full name of each cell type specifying the species, for example 'am' for Apis mellifera.

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

We are indebted to the many test users of the IBdb who patiently located bugs and inconsistencies and thereby helped to streamline the database outline and make the user interface more intuitive. We also thank all members of the Heinze lab for many helpful discussions that improved the IBdb and this manuscript.

© 2021, Heinze et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.