InsectBrainDatabase - A unified platform to manage, share, and archive morphological and functional data

Insect neuroscience generates vast amounts of highly diverse data, of which only a small fraction are findable, accessible and reusable, despite open data mandates by funding bodies. We have therefore developed the InsectBrainDatabase (IBdb), an open platform for depositing, sharing and managing a wide range of insect neuroanatomical and functional data. It facilities biological insight by enabling effective cross-species comparisons, by intimately linking data on structure and function, and by serving as hub for information on insect neuroethology. The IBdb provides novel visualization and search tools, which are also available in a unique private mode of the database, before data is made public. This allows users to manage and visualize unpublished data, creating a strong incentive for data contribution and eliminating additional effort when publicly depositing the data at a later stage. These design principles could also serve as a blueprint for similar databases in other fields.


31
Data are the essence of what science delivers -to society, to researchers, to engineers, to en-32 trepreneurs. These data enable progress, as they provide the basis on which new experiments Elements in each layer are represented on their respective profile pages, which can be located either through lists or search interfaces. B. Examples for information that can be deposited on each level of the database, illustrating how diverse data is automatically associated with hierarchical metadata. C. Schematic illustration of a neuron profile page. Detailed anatomical and functional data is available on this page, from where experiments associated with this cell type are also linked. Similar pages exist for species and brain regions. 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 140 persistent.

141
Interactive search interfaces 142 As locating specific datasets is one of the core functions of a database, we have developed a novel, 143 more intuitive way to find specific neuron data. A graphical representation of the insect brain, re-144 sembling the overall anatomical outline of all brain regions (Figure 2A), makes it easy to search for 145 neurons within single species and across species. This graphical interface is generated directly on 146 the database website for each species and is adapted from a generic insect brain, i.e. a shared 147 ground plan. This generic brain is the least detailed fall-back option for any cross-species search 148 and was developed based on the insect brain nomenclature developed by Ito et al. (2014). It resem-149 bles the consensus anatomical hierarchy of all brain regions in insect brains. Within this hierarchy, 150 the entire brain is divided into 13 super-regions, which consist of individual neuropils. The lat-151 ter can be further divided into sub-regions. While all super-regions exist in all species, differences 152 become more pronounced at lower levels of the hierarchy. The generic brain therefore largely con- As these categories are simply tags of brain region entries used to 155 organize the database, the search interface does not differentiate between them, simplifying the 156 user experience (Figure 2A). 157 If more than one species is subject to a query, a schematic brain is generated that displays 158 the commonly shared features of the species involved. For both single and multi-species search, 159 when selecting a specific brain region, all neurons in the IBdb that connect to this region become 160 visualized by a dynamically drawn wiring diagram ( Figure 2B). Filters can be applied to narrow down 161 search results according to neuron polarity, functional class, etc. Individual neurons in the wiring 162 diagram can be selected to reveal the neuron's profile page. Here, all available information for this 163 cell type is displayed, including links to deposited experiment entries ( Figure 2C). The schematic 164 display of search results can visualize any neuron in the database, only requiring that a neuron is 165 annotated with respect to the brain regions it innervates. 166 For single species queries, two more modes for visualizing search results are available: the 167 semi-schematic view and the 3D view. The semi-schematic view mode emphasizes the natural 168 brain organization on the level of brain regions, while also serving as interface for launching search 169 queries. It comprises a full series of automatically generated sections through a segmented 3D 170 brain of a species. Each brain region present in that species' 3D brain is shown as an interactive 171 cross section that can be used to query neural connections of that region (Figure 2D). If a region is 172 selected, all connected brain areas are highlighted and neurons resulting from this query can be 173 visualized by changing to either the schematic view, the 3D view, or a list view. The advantage of the 174 anatomically correct layout of this interface is that a brain region can be queried for a neuron, even 175 if its name is not known to the researcher. This is particularly useful for regions with uncommon 176 names that have only recently been introduced to the insect brain naming scheme (e.g. crepine,  The 3D view visualizes search results in an anatomically correct way and shows queried neurons 183 in the context of a species' reference brain (given that this information was added) (Figure 2E). It 184 displays interactive surface models of that brain together with neuron skeletons obtained from 185 the neuron-type's profile page.

186
In the graphical search interface the search parameters are limited to anatomical information 187 defining the neuron's location in the brain (i.e. input and output areas). In contrast, an additional 188 Figure 2. Neuron search in the IBdb. A. Screenshot of schematic brain search interface (Monarch butterfly) in single species mode. Selecting a neuropil will reveal all neurons connected to that neuropil. B. Schematic wiring diagram view of search result; orange neuropil was queried. Selecting an individual neuron will reveal the profile page of that cell type. C. Example of neuron profile page with anatomical information. Confocal image stack and functional data are not listed in this example. D. Semi-schematic search interface. The section view is scrollable and allows the user to query individual neuropils for connected neurons by clicking the cross section. The inserts show the results view at three levels of the brain. Neuropils connected to the queried neuropil are highlighted. Switching to the schematic view will then show the neurons as wiring diagram. Switching to the 3D mode will show registered neurons in 3D. E. The 3D results viewer allows one to view all neurons registered into a common reference frame; example from Monarch butterfly (data from Heinze and Reppert (2012)) . The user can continuously switch between the three modes (schematic, semi-schematic, 3D).    193 Finally, whereas the emphasis of the database search lies on locating cell types, information 194 on brain regions can also be found using identical interfaces. The schematic search option allows 195 users to reveal brain region profile pages by selecting schematic neuropil representations. The 196 same information can also be obtained by selecting brain regions in the semi-schematic neuropil 197 search interface. Due to their heterogeneous nature, experiment entries can only be queried via 198 the text-based search function ('Expert Search').

199
Online applications and tools 200 To maximize the usefulness of the database we have implemented an integrated 3D viewer to 201 deliver platform independent, high-quality data visualization without any additional software de-202 mands. Neuropil visibility can be independently switched on and off for each brain region, trans-203 parency can be freely adjusted, and colors of neurons can be changed ( Figure 3A). Neurons can be 204 shown either with diameter information or as simple backbones. The built-in screenshot function 205 enables the user to capture any scene displayed in the 3D viewer and produces a high-resolution, 206 publication-ready image with transparent background (Figure 3B,C). 207 The IBdb allows users to not only locate neuronal morphologies quickly, but also to combine 208 arbitrary neurons from any single species into a common visualization. To achieve this, we have 209 generated a neuron clipboard, in which individual neurons from search results can be stored tem-210 porarily ( Figure 3D). Any subset of cells in the clipboard can be sent to the 3D viewer, as long as 211 all neurons belong to the same species, i.e. can be displayed using the same reference brain. The 212 desired configuration of neurons and neuropils can be generated using the interactive tools of the 213 viewer and the screenshot function can be used to create a high-resolution image to be used for 214 illustration purposes (e.g. reviews, conference talks, teaching).  four-window 3D viewer retains all functions of the normal full screen 3D viewer and thus also allows 221 the capture of high resolution screen shots of each of the neurons being compared ( Figure 3E). 222 The data in the database are suited for many applications, including more sophisticated ones.

223
To provide direct access to all levels of the data in the IBdb we have created an API interface,

Contribution of data
All data on the internet is public. This also applies to any data publicly available in the IBdb. Driven 237 by the requirement to obtain data persistency and implemented by the use of handles, no data 238 can be removed from the database once it is public (and thus citable). For all data, the contributor 239 retains ownership and holds the copyright to her/his data. The publishing is performed explicitly To create a new species, a profile page has to be created, which then has to be populated with data.

253
While photographs, bibliography, and text descriptions of the species are desirable and strongly 254 encouraged, the most important next step is to generate a schematic brain ( Figure 4A). This will 255 ensure that brain structure entries are created in the database, a prerequisite for neuropil based 256 search and 3D brain region identity. To generate a schematic brain, we have created the 'Brain 257 Builder' tool on the IBdb website, which provides templates based on either the generic insect 258 brain, or related species already deposited in the database. The user can simply copy an existing 259 brain, associate it with the new species and modify it to match any unique features of the new 260 species.

261
Once the profile page and the schematic brain is created, a 3D brain can be uploaded to illus-262 trate the brain organization of the new species and to serve as reference brain for neuron display 263 ( Figure 4A). For all uploaded data, the database distinguishes between source files and display files.

264
Source files contain the 3D reconstruction in a format that the researcher would like to make avail-265 able for others in the field. The second set of files, the display files, are required for automatic 266 online display and must constitute the surface models of each neuropil (.obj-format). Each brain 267 region model is tagged with a unique neuropil identity, so that the schematic brain regions and 268 the 3D surface models will be linked to the identical brain-structure entry. Finally, following the 269 same principles as for the 3D brain, an image stack can be uploaded to the profile page as well 270 ( Figure 4A). This can be any representative dataset that illustrates the layout of the species' brain 271 (e.g. confocal stack, µCT image series, serial sections with any other technique).

272
Brain structures 273 Brain structure entries are automatically generated when defining the schematic brain search in-274 terface for a new species ( Figure 4B). These profile pages are automatically populated by a 3D brain 275 in which the relevant region is highlighted, a brain structure tree that reveals the relative location 276 of the respective neuropil in the hierarchy of the species' brain, and with links to neuron entries 277 associated with each brain region. All remaining data have to be manually added. These are mostly 278 descriptive in nature and encompass images, text-based descriptions, and volumetric data.

279
Neurons 280 Contribution of neurons follows a similar procedure as the contribution of new species (Figure 5). 281 The user creates a profile page for the new cell type that subsequently has to be populated with 282 information. To make a neuron findable in the database, its arborization regions in the brain have 283 Figure 4. Contributing a species to the IBdb. A. Three main elements have to be created for each new species: the schematic brain, the 3D brain and an image stack. The schematic brain is generated directly on the IBdb website using the 'Brain Builder', while the other two elements are uploaded. For each, both source files and viewer files are needed. Viewer files are used for online display, while source files can be downloaded by users. B. Neuropil profile pages are automatically generated when creating the schematic brain. They have to be populated with images and texts by the user. C. The semi-schematic brain is automatically generated based on the provided 3D brain. D. The species profile page must be populated with images, texts and a bibliography to provide context for the species. Photograph reproduced with permission from Ajay Narendra. This generates a neuron profile page that then has to be populated with information by the owner. Each entry is findable by the expanded search function. B. 3D-skeletons are added as swc-files (online display) and source files (download). Confocal image stacks are uploaded as jpeg series for online display and as original data files for download. C. All arborization regions of the neuron must be defined (at the level specified in the species' schematic brain) and labeled as either input, output or unknown polarity. D. To enable automatic drawing of wiring diagrams in schematic search results, the order of innervation of neuropils and the branch-points of the neuron must be defined using the path assistant.
to be defined (Figure 5C). Within a graphical user interface, these regions are chosen from the 284 brain structures available in the schematic brain of the respective species. One arborization entry 285 has to be created for each branching domain of the neuron, leaving no part of the neuron un-286 annotated. To enable the automatic generation of a wiring diagram view of the new neuron for 287 displaying schematic search results, an outline of its branching structure has to be generated in 288 an embedded tool called the neuron-path assistant ( Figure 5D). This branch tree defines which 289 neuropils are innervated in which order and where main branch points are located.

290
The remaining procedure for neuron contribution is largely identical to species contribution 291 and follows the dual approach towards source data and display data for 3D reconstruction and 292 image stacks ( Figure 5B). All other information, i.e. images, bibliography, keywords, representative 293 functional data, transmitter content, and textual descriptions, can be added to the profile page at 294 any time prior to publication. 295 Importantly, the IBdb can be used to house data that have been obtained by classical methods, 296 e.g. camera lucida drawings of Golgi impregnated neurons. While no 3D information is available 297 in those cases, drawings can be uploaded as images after which annotation of the neuron's mor-298 phology (arborization regions) is performed as described. These neurons will therefore become 299 findable in the schematic search interface and will be added to the publicly available pool of neu-300 ronal data.

301
Experiments 302 Experiment entries are created by adding them directly to the profile page to which they are linked 303 (species, brain structure, or neuron). The automatically generated experiment profile page must 304 then be filled with basic meta-information about the experiment (date, what was done, who did it), after which a series of files can be uploaded. These files can be in any format and are made 306 available for download. This allows users to provide not only the raw data of any experiment, but 307 also, for example, analysis scripts, custom made equipment-control software, and analysis results.

308
Image files can be selected for direct online display on the experiment profile page to allow online 309 examination of the data.

310
Curation and administration 311 The database is managed via a group of voluntary curators, a scientific administrator, and a tech-312 nical administrator. Importantly, no single person curates all data in the IBdb, but each species is 313 managed by a specialized curator, who is an expert for that species. This distributed curation sys-314 tem ensures that no single person is responsible for too many datasets, and that no curator has 315 to evaluate data outside their area of expertise. To additionally reduce the workload for species 316 with many entries, more than one curator can be assigned to any given species. The scientific 317 administrator (the lead author of this publication) oversees the curators, while technical adminis-318 tration is carried out by the technical administrator (last author of this publication). The technical 319 administrator is the only person who has potential access to all data in the database.  The owner can then directly respond to the comments and any issues raised can be resolved.

329
To enable all users to provide feedback and to discuss topics relevant to other database users, 330 we have added a discussion forum directly to the IBdb website. This forum is intended as a means 331 for reporting potential bugs, suggesting new IBdb features, or for discussing scientific content 332 (methods for data processing or acquisition, requests for literature, staining protocols etc.).

333
The IBdb as tool for data management and data deposition 334 Each database entry has to be explicitly published by the contributor. In the process, it is approved 335 by either the database administrator (species), or the species' curator (neurons). While this proce-336 dure was initially intended only as a quality control measure to prevent incomplete or inaccurate 337 data from compromising the database, we have developed it into a unique feature: the IBdb private 338 mode. Before a dataset is made public, it is invisible to all other users, curators and the scientific 339 database administrator. The dataset can thus be updated and even deleted. This creates the po-340 tential of using the IBdb to deposit data while they are being collected or prepared for publication 341 in a research paper, i.e. for data management (Figure 6A). 342 To facilitate the use of the IBdb as a data management tool, we have enabled three operational 343 modes of the database site: private, public, and mixed. Any user logged in can thus choose to 344 either access only (own) private data, only public data, or both. The first mode turns the IBdb into 345 a data management site for ongoing research, the second mode is the default mode for viewing 346 publicly available data, and the third mode allows the user to compare their own unpublished data 347 with public data.

348
As efficient data management requires researchers from the same laboratory, as well as col-349 laborators, to have access to relevant unpublished data, each user can grant access to their own 350 private datasets (Figure 7). To this end, a user can create a user group and invite other database 351 users to join. Datasets can then be added and made visible or editable to all members of the group.  Finally, users are often reluctant to make datasets available to the public before they are in-355 cluded in a research paper, yet, these data should be available to editors and anonymous review-356 ers. We have therefore created the possibility for 'pre-publishing' database entries (Figure 6A). This  To avoid having to separately provide numerous independent handles when sharing data, indi-364 vidual entries can be grouped into datasets. These receive a unique link that grants access to the 365 entire collection. Only public or pre-public data can be included in datasets.

366
Public entries of the IBdb are maintained in a dual way; the persistent version is locked and 367 cannot be changed, whereas a second, current version remains visible to the owner and to all 368 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 370 version can be deleted. Rather, when for instance a confocal image stack should be replaced by 371 a better one, the old stack can be archived, so that it will not be visible in new versions of the 372 dataset, but will remain present in the database for display of earlier versions. Once all required 373 updates of a dataset have been made, the edited version can be re-published and will be assigned   Figure 7. The Insect Brain Database and the possible interactions between users and deposited data. The private sections of the database are accessible to only the owner of the data, and datasets within this section can be shared with team members and collaborators. As these datasets are unlocked, they can continuously be updated and also deleted. Upon publication and curator approval, datasets become locked (persistent) and are archived in the public section of the database. As an intermediate step, datasets can be pre-published (locked but private) and made available to journal editors and peer reviewers when including datasets in manuscripts of journal articles. Data in the public section of the database is accessible directly for all interested users (relevant user groups are shown on the right). Additionally, an application programming interface (API) allows automated access of public data, which can therefore be used by third party applications (illustrated as 'App 1-3') for generating specific user experiences with additional capabilities, for instance in the context of teaching. of larger databases was generally impaired by a layout that often required expert knowledge to 389 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 391 old databases, but the restriction to individual species and often highly complex interfaces limit 392 the potential user base to specialists even in cutting edge databases such as VFB or visualization 393 tools such as FruitFlyBrainObservatory. Finally, the limitation to purely anatomical data, including 394 in the major current cross-species database NeuroMorpho.org, does not account for one of the 395 key advantages of insect neuroscience: the high level of tractable structure-function relations.

396
The IBdb addresses each of these issues. Firstly, we have developed the database software to 397 be independent of the operating system and type of web browser used, as well as to not rely on 398 any third party plugins. Additionally, we implemented the database as a classic, relational database 399 without experimental data structures (e.g. intelligent, adaptive search), aiming at maximal robust- and functional data additionally broadens the relevance of the IBdb. 437 We have implemented a range of tools enabling the visualization of data in fast, flexible, and ef-438 fortless ways. This saves considerable time compared to other available data visualization tools, in 439 particular for complex 3D neuron data (e.g. Amira). The data contributed are also immediately in-440 corporated into the framework of existing data. Outside the IBdb these data are distributed across 441 many publications. Comparison of one's own data to any published data would entail contacting 442 authors, obtaining files in unpredictable formats and finding ways to compare them to one's own 443 work within the software a research group is currently using. The IBdb solves these issues and 444 delivers such comparisons within seconds. 445 Crucially, these advantages are already present immediately after data upload, prior to pub- groups that initiated the database have paid for its creation. As the maintenance costs are a small 459 fraction of the development costs, it will be easily possible to run the database within the frame-460 work of the existing service agreement for at least the next five years without any changes required.

461
However, when the data volume increases substantially, the static costs of housing the data will 462 increase accordingly. While keeping all public, persistent data available free of charge is manda-463 tory (given that the IBdb functions as a public data repository), maintaining the IBdb as a free data 464 management tool, i.e. allowing unlimited private data for each user, will likely become unsustain-465 able over time. If this becomes a problem, free space in the private section of the database will be 466 limited. All space required beyond a certain limit will have to be rented to directly offset the costs 467 for maintaining and administering these data.

468
To anticipate the slowly growing costs of housing the database due to increasing data volume, 469 we aim at eventually relocating the data from the currently used commercial Amazon cloud plat-470 form to an academic server that is provided at minimal costs or free of charge. To this end we have 471 ensured that the IBdb does not depend on any core functions of the Amazon cloud storage service, 472 enabling to move the database to a new location with comparably moderate effort.  Finally, the IBdb provides the possibility for anyone to access original research data in intuitive 483 and attractive ways (Figure 7). This provides opportunities to design teaching assignments for 484 neuroscience students to carry out meta-analyses. With access to the data in the IBdb via the ap-485 plication programming interface (API), we have provided the possibility for third parties to develop 486 dedicated teaching tools that provide streamlined methods to use the data for specific classroom 487 exercises. Beyond researchers and students, journalists, interested members of the public, or 488 members of funding bodies can also view and explore neuroscience data. Ideally, this will con-489 tribute to a more transparent understanding of what the output of science is and could spark 490 increased interest in insect neuroscience.

492
The IBdb does intentionally not include Drosophila melanogaster as a species. This is because a huge 493 amount of effort has been spent developing highly efficient resources for this widely used model  (Scheffer et al., 2020). Serving as the main repository for anatomi-497 cal data from the Drosophila brain it has become the main site to locate GAL4 driver lines, single 498 cell morphologies, and synaptic connectivity data. It contains tens of thousands of datasets and 499 is designed to specifically meet the needs of the Drosophila research community. By being less 500 specialized, the IBdb has a wider scope. We are hosting many species and include both functional 501 and anatomical data. We also do not require neuronal anatomies to be registered to a reference 502 brain, if this is not possible for some reason. This opens the IBdb up to more diverse data, but as a 503 result cannot provide most of the specialized services that VFB can deliver (e.g. automatic bridging 504 registrations of 3D data between different reference brains). Importantly, both databases have 505 converged on a highly similar ontological framework. As the brain nomenclature used by the IBdb 506 and VFB is identical, neuropil identities can be mapped across both databases and direct links can 507 be provided between them. Ideally, this will eventually enable a user to launch a query in the IBdb 508 and directly link the results to the corresponding data in VFB (and vice versa). This interoperability 509 will make maximal use of both complementary resources, without duplicating functionality.

510
Widening the scope towards other animal groups 511 The framework we have generated with the IBdb is not limited to housing insect brain data. Without 512 major modifications it would be equally suited for hosting data from other animal taxa. While the 513 intuitive, schematic search engine would not be useful for comparing species that do not share a 514 common basic brain outline (i.e. a relevant 'generic brain'), the text-based expanded search could 515 allow the construction of queries across multiple groups, e.g. searches according to functional 516 terms. We are currently conceptualizing the expansion of the database towards including spiders 517 and envision that crustaceans and other arthropods would be logical next groups.

518
The IBdb therefore provides not only a tool for the insect neuroscience community to facilitate 519 data management, data visualization, transparency of results and effective teaching, but it can 520 easily be expanded towards related fields. Additionally, it might also serve as a blueprint for how 521 to set up similar databases in unrelated research areas. In principle, the strategies used in the IBdb 522 are applicable to any scientific field that can be linked to a hierarchical, ontological framework.

Data location 525
All web infrastructure is hosted by Amazon web services on servers located in Frankfurt, Germany.

526
Data is stored using a PostgreSQL relational database hosted by the Amazon Relational Database 527 Service and files are stored using the Amazon S3 object storage service. The servers hosting the 528 website and the local HANDLE system are running in Amazon EC2 containers, which runs Linux.

529
Resources communicate using Amazon Virtual Private Cloud.

530
Database framework 531 The database structure and interaction is managed by a python based Django application. User 532 authentication, permissions and data security are also managed within the Django application. A 533 NGINX web server hosts static content and serves as a reverse proxy for dynamic content served 534 by a uWSGI application server hosting the project's Django application. Asynchronous tasks are 535 implemented using the Celery distributed task queue and RabbitMQ message broker.

536
Data is externally accessible via a web API delivering content in the JSON format to the front-end 537 web application. The web API was implemented with Django and the Django REST framework.

538
Long term data persistency is provided to allow users to reference information or profile pages 539 on the site in scientific publications and other external media in a static state, while continuing to 540 allow data to be updated as more information is acquired. When a request is made by a user for 541 a persistent copy of a dataset to be created, a copy of the data related to the current state of the 542 dataset is serialized and parsed into JSON. A persistent unique identifier is then assigned (HANDLE).

543
The JSON data, HANDLE and additional metadata is recorded in a separate table and can no longer 544 be modified. All files associated with the persistent dataset are marked as locked in the database 545 and can no longer be modified by the user. The recorded state of that dataset can be accessed and 546 viewed on the site using the url associated with the assigned HANDLE. The original data copied to 547 create the persistent dataset can be modified without effecting the persistent dataset. Additional 548 files may also be added, but will not be reflected in earlier persistent records. and Gecko based layout engines adhering to web standards.

555
The web based three-dimensional viewer was implemented using Typescript, WebGL and the 556 Three.js three dimensional graphics framework. The two-dimensional schematic view, brain de-557 signer and path designer was implemented using Typescript, the Canvas API and Paper.js vector 558 graphics scripting framework.   are often specific to individual species and therefore, if such regions were defined, we used the 587 names given to them within the relevant species. We did not unify e.g. names of the mushroom 588 body calyx divisions across species, as this would firstly imply homology where there might not be 589 any, and second, novel naming schemes will have to be developed by the community and not be 590 imposed by a data repository. Anticipating that changes to brain names can and will happen in the 591 future, all names, as well as the level of a region within a hierarchy, can be modified.

592
Within some neuropils, regular, repeating elements can be found, usually defined as columns 593 and layers. We have implemented such a system in the central complex, i.e. without having to de-594 fine an array of sub-regions, several strata and orthogonal slices (following the new brain nomen-595 clature) can be generated. The default number of slices in the generic brain is 16, assuming that 596 this number is the ancestral state of this region.

597
Neuron names follow the conventions within each species, as there is no common naming 598 scheme for insect brain neurons yet in place. However, we provide the possibility to define several 599 alternative names for each cell type to allow the parallel use of names. This is possible as the 600 identity of a neuron is linked to the persistent ID, and not to the neuron's name. Given that we 601 house neurons from multiple species, we add a prefix to the full name of each cell type specifying 602 the species, e.g. 'am' for Apis mellifera.

603
Acknowledgments 604 We are indebted to the many test users of the IBdb who patiently located bugs and inconsistencies 605 and thereby helped to streamline the database outline and make the user interface more intuitive. 606 We also would like to thank all members of the Heinze lab for many helpful discussions that im-    The brain builder module enables to generate the schematic brain models required for the graphical search functions. It allows the user to draw brain regions (tagged as either super-region, neuropil, or sub-region) as well as fiber bundles and invisible bridges. The latter provide shortcuts for neurons in the wiring diagram view (schematic search results), preventing overly convoluted neuron paths. The top panel shows the generic brain that is used as the template for all other species and which combines brain regions and fiber bundles shared by the majority of insect species. The bottom four panels illustrate the modifications of this template needed to account for species-specific features of four insect species examples. Brain structures included in the insect brain database. All regions shown in black font are included in the generic insect brain that serves as a template for all other species and as a fall-back option for cross-species search. Bold regions are super-regions, which consist of neuropils, which in turn can be divided into sub-regions. Note that neuropils and subregions can be species-specific (grey font) and no single species contains all listed structures.