BRIDE v2: A Validated Collection of Genes Involved in the Mammalian Brain Response to Low-Dose Ionizing Radiation

There is significant interest in the response of the mammalian brain to low-dose ionizing radiation (LDIR), mainly examined by gene or protein expression, with applications in radiation safety on Earth, the atmosphere and outer space. Potential associations of molecular-level responses with sensory or cognitive defects and neurodegenerative diseases are currently under investigation. Previously, we have described a light-weight approach for the storage, analysis and distribution of relevant datasets, with the platform BRIDE. We have re-implemented the platform as BRIDE v2 on the cloud, using the bioinformatics infrastructure ELIXIR. We connected the annotated list of 3174 unique gene records with modern omics resources for downstream computational analysis. BRIDE v2 is a cloud-based platform with capabilities that enable researchers to extract, analyze, visualize as well as export the gene collection. The resource is freely available online at <http://bride-db.eu>.


Introduction
Exposure to low-dose ionizing radiation (LDIR) is an important public health factor, with long-term consequences for human well-being [1]. Common situations for LDIR exposure relate to medical diagnostic procedures [2] and frequent air travel [3]. More unusual settings involve specific professional groups and range anywhere from nuclear power stations to space travel [4]. It is estimated that our annual LDIR exposure has increased in just one generation six-fold, from 0.5mSv in 1980 to 3mSv today [5]. Particular emphasis has been given to the molecular mechanisms of LDIR response beyond dosimetry [6]. With omics technologies, it is expected that we can better understand the genetic variations or expression patterns relevant for LDIR response, to improve radiation protection [7]. Earlier considerations can now be addressed with precise measurements at the genomic, transcriptomic and proteomic levels [8]. The planning of long-haul space flight missions has also invigorated the field and spawned renewed interest [9].
Multiple systems have been profiled with omics technologies, including human skin [10], heart tissue [11] and stem cells [12]. Of primary concern, however, is the effect of LDIR on the brain, both for impairing cognitive function and the long-term consequences Radiation 2022, 2 312 such as neurodegenerative disease [13]. Identifying molecular markers that can be used as proxies for dosimetry is also a key factor in this research, in particular the mammalian brain [14]. Beyond gene expression responses and the quest for potential protein biomarkers, microRNAs might also be involved [15], as well as various epigenetic effects [16]. Despite progress, omics profiling of LDIR in mammalian tissues is still at its infancy [17].
As little is known about the cerebrovascular system and the brain, it is important to record gene information for brain-specific LDIR response within a broader context, to validate existing data and guide future experimentation. The BRIDE resource was launched six years ago, with rather limited infrastructure [18]. We now relaunch the original collection as BRIDE v2 on the cloud, with additional and more modern tools that can facilitate research into this domain. BRIDE v2 follows the initial philosophy of BRIDE [18], in particular a gene-centric view for the curation of relevant genes found to respond to LDIR in the mammalian brain.

Design and Content Overview
We have re-implemented BRIDE following the same design principles as in v1 [18], with a slight update of the original tabs and in particular the PC-Viz tool, now replaced by a GeneMania client call [19]. We opted for a light-weight approach to data integration [20] as previously reported [18], using responsive design principles for a variety of screens and devices. The documentation for each entry has been extended and now includes additional fields ( Figure 1). While the collection does not represent by itself an instance of the big data domain [21], the implied and/or extended datasets for each gene record are significant; 3174 unique records, with an average of 15-20 links per record, the collection points to over 50,000 static or dynamic data frames.
as proxies for dosimetry is also a key factor in this research, in particular the mamm brain [14]. Beyond gene expression responses and the quest for potential prote omarkers, microRNAs might also be involved [15], as well as various epigenetic e [16]. Despite progress, omics profiling of LDIR in mammalian tissues is still at its in [17].
As little is known about the cerebrovascular system and the brain, it is import record gene information for brain-specific LDIR response within a broader context, t idate existing data and guide future experimentation. The BRIDE resource was laun six years ago, with rather limited infrastructure [18]. We now relaunch the original c tion as BRIDE v2 on the cloud, with additional and more modern tools that can fac research into this domain. BRIDE v2 follows the initial philosophy of BRIDE [18], in ticular a gene-centric view for the curation of relevant genes found to respond to LD the mammalian brain.

Design and Content Overview
We have re-implemented BRIDE following the same design principles as in v1 with a slight update of the original tabs and in particular the PC-Viz tool, now rep by a GeneMania client call [19]. We opted for a light-weight approach to data integr [20] as previously reported [18], using responsive design principles for a variety of sc and devices. The documentation for each entry has been extended and now include ditional fields ( Figure 1). While the collection does not represent by itself an instan the big data domain [21], the implied and/or extended datasets for each gene recor significant; 3174 unique records, with an average of 15-20 links per record, the colle points to over 50,000 static or dynamic data frames. The basic BRIDE data front-end design has not been significantly altered, main ing the unification and relationship links for the detection of the identity and associa of each gene, respectively [22], yet the actual tools that serve those queries have slightly modified, as well as the underlying, back-end architecture (see section 2.2 example, the PathwayCommons tool PC-Viz has been replaced by GeneMania [19] more convenient and comprehensive solution. Dose-time records are also provid each case, as well as tissue and organism (species). The effort aims at further suppo activities for systems biology and omics experiments [23] in the field of LDIR mamm brain response. The basic BRIDE data front-end design has not been significantly altered, maintaining the unification and relationship links for the detection of the identity and associations of each gene, respectively [22], yet the actual tools that serve those queries have been slightly modified, as well as the underlying, back-end architecture (see Section 2.2). For example, the PathwayCommons tool PC-Viz has been replaced by GeneMania [19], as a more convenient and comprehensive solution. Dose-time records are also provided in each case, as well as tissue and organism (species). The effort aims at further supporting activities for systems biology and omics experiments [23] in the field of LDIR mammalian brain response.

Design Details
As stated previously [18], we have compiled information for genes involved in the mammalian response to LDIR response, validated from the literature and omics results from the CEREBRAD project. The unification links [22] connect gene records to the UniProt knowledge base [24], both for identifiers and accession numbers. Links to the relevant publications are also provided at the main entry point. Using the NCBI tab, more links including relationship links [22] are given for genomic location, pathways [discontinued in March 2022] and gene expression profiles [25]. The same strategy is used for the IntAct database of protein interactions [26] and a selection function for GeneMania [19]. Finally, rich links to relevant RBStore experiment-centric records, the Allen Mouse Atlas [27] anatomy and gene expression records and MGI phenotypes, where available, [28] are also provided for completeness ( Figure 1).

Content Details
In the v2 re-implementation, some unification links have been redesigned, as data providers have changed the API details; we now reproduce the entire dataset through the new APIs to ensure proper connection with external databases. The data have been imported into a PostgreSQL database. All user queries and choices target the database which returns results through a web browser. BRIDE v2 supports access via all modern browsers and mobile devices, without the need of any plugin installation. The full dataset is available in JSON format for download by end users-seen at the bottom of the page at all levels of access or tabs. The open nature of the dataset encourages potential extensions that may capture additional details from the original publications, such as type of irradiation, linear energy transfer (LET) or phenotypic information.

System Architecture
The BRIDE v2 platform implementation is based exclusively on AWS cloud services, in contrast to v1, which was based on a 3-tier architecture with MS ® Azure Cloud and SQL [18] (no longer available). The services for this implementation and the system architecture are shown (Figure 2). The design philosophy of the system is based on the serverless computing model, with AWS Lambda which has revolutionized the application landscape. The function-as-a-service approach lets users deploy cloud applications without the need of a 24/7 active server. Instead of hosting an entire application on a server, we simply need to upload an individual function to AWS Lambda and call it using one of the many triggers available to initiate a function call.
iation 2022, 2, FOR PEER REVIEW 2.1.1. Design Details As stated previously [18], we have compiled information for genes involved in mammalian response to LDIR response, validated from the literature and omics resu from the CEREBRAD project. The unification links [22] connect gene records to the U Prot knowledge base [24], both for identifiers and accession numbers. Links to the relev publications are also provided at the main entry point. Using the NCBI tab, more li including relationship links [22] are given for genomic location, pathways [discontinu in March 2022] and gene expression profiles [25]. The same strategy is used for the Int database of protein interactions [26] and a selection function for GeneMania [19]. Fina rich links to relevant RBStore experiment-centric records, the Allen Mouse Atlas [27] an omy and gene expression records and MGI phenotypes, where available, [28] are also p vided for completeness (Figure 1).

Content Details
In the v2 re-implementation, some unification links have been redesigned, as d providers have changed the API details; we now reproduce the entire dataset through new APIs to ensure proper connection with external databases. The data have been ported into a PostgreSQL database. All user queries and choices target the database wh returns results through a web browser. BRIDE v2 supports access via all modern brows and mobile devices, without the need of any plugin installation. The full dataset is av able in JSON format for download by end users-seen at the bottom of the page at levels of access or tabs. The open nature of the dataset encourages potential extensi that may capture additional details from the original publications, such as type of irra ation, linear energy transfer (LET) or phenotypic information.

System Architecture
The BRIDE v2 platform implementation is based exclusively on AWS cloud servi in contrast to v1, which was based on a 3-tier architecture with MS ® Azure Cloud and S [18] (no longer available). The services for this implementation and the system archi ture are shown ( Figure 2). The design philosophy of the system is based on the serverl computing model, with AWS Lambda which has revolutionized the application la scape. The function-as-a-service approach lets users deploy cloud applications with the need of a 24/7 active server. Instead of hosting an entire application on a server, simply need to upload an individual function to AWS Lambda and call it using one of many triggers available to initiate a function call. AWS Lambda provides a stand-alone execution environment for individual fu tions written in Python. Thus, we can significantly reduce the total costs and the hardw maintenance needs. Unfortunately, the programmer has no control over the environme and needs to manipulate more complex patterns to ensure proper functionality, spec AWS Lambda provides a stand-alone execution environment for individual functions written in Python. Thus, we can significantly reduce the total costs and the hardware maintenance needs. Unfortunately, the programmer has no control over the environment, and needs to manipulate more complex patterns to ensure proper functionality, specifically orchestrating and organizing the functions, so that they can work in a distributed fashion on available data. AWS supports PostgreSQL as a fully managed database service with Amazon Relational Database Service (RDS). PostgreSQL is an open-source database that supports both SQL and JSON queries. The Amazon API Gateway is a fully managed service to create, publish, maintain, monitor, and secure APIs at any scale. APIs act as the 'front door' for applications to access data, business logic or functionality from back-end services. This API easily accepts and processes up to hundreds of thousands of concurrent API calls and is very well suited for integration with AWS Lambda. The BRIDE v2 front-end has been developed using jsGrid, which is a lightweight client-side data grid control based on jQuery. It supports basic grid operations like inserting, filtering, editing, deleting, paging and sorting. jsGrid is flexible and allows customization of its appearance and components. With this approach, we ensure a minimal interface which presents all the available information without the need for complex navigation and wasteful resource management.

Data Consumption
Users can access records in BRIDE v2 with simple text queries, sort functionality and other search capabilities, e.g., a combination of the above, as following: first filter by text, then sort entries. Examples of the interface are provided in a pictorial manner (Figure 3).
Radiation 2022, 2, FOR PEER REVIEW 4 with Amazon Relational Database Service (RDS). PostgreSQL is an open-source database that supports both SQL and JSON queries. The Amazon API Gateway is a fully managed service to create, publish, maintain, monitor, and secure APIs at any scale. APIs act as the 'front door' for applications to access data, business logic or functionality from back-end services. This API easily accepts and processes up to hundreds of thousands of concurrent API calls and is very well suited for integration with AWS Lambda. The BRIDE v2 frontend has been developed using jsGrid, which is a lightweight client-side data grid control based on jQuery. It supports basic grid operations like inserting, filtering, editing, deleting, paging and sorting. jsGrid is flexible and allows customization of its appearance and components. With this approach, we ensure a minimal interface which presents all the available information without the need for complex navigation and wasteful resource management.

Data Consumption
Users can access records in BRIDE v2 with simple text queries, sort functionality and other search capabilities, e.g., a combination of the above, as following: first filter by text, then sort entries. Examples of the interface are provided in a pictorial manner (Figure 3). At all levels, the download functionality is provided at the bottom of any page. The BRIDE v2 collection is identical to the previously published list, namely 3174 records, using gene names as handles, plus the tissue-dose-time tuple (after irradiation). In other words, time refers to the time-period from irradiation to measurement. As previously reported, 95% of entries correspond to protein-coding genes expressed in the brain The BRIDE v2 collection is identical to the previously published list, namely 3174 records, using gene names as handles, plus the tissue-dose-time tuple (after irradiation). In other words, time refers to the time-period from irradiation to measurement. As previously reported, 95% of entries correspond to protein-coding genes expressed in the brain of Mus musculus. The additional genes are kept as they are reported in the same studies, documented by the corresponding PubMed identifier (PMID). Amongst the records that derive from four major publications, there are also a small number of unpublished instances (154 in total). Manually curated records (165 in total) are also included [18], and further curation efforts are envisaged.
With~16 links per record, there are over 50,000 active links to other tools and databases (Table 1), thus creating a dense network of connections from BRIDE to some of ELIXIR's core data resources [29] and others. Users are able to generate dynamic links with the Genemania tool [19]: more than one gene can (and should) be added to the query string and submitted to Genemania, in order to predict functional associations of the selected genes. Of the total number of entries, 1965 (62%) refer to radiation doses ≤1Gy, thus making the collection particularly relevant for LDIR research.

Utility and Discussion
The re-implementation of BRIDE as a new version provides an updated resource for relevant genes known to be affected by LDIR in the mouse brain and a toolkit for further analyses. We have upgraded the previous design due to changes both in the data content/specifications as well as the plugin-free serverless architecture options that are now available. We maintain the choices for platform usability and portal access and make the entire dataset available to the scientific community in JSON format. Further integration with other relevant resources and better connection to ELIXIR bioinformatics platforms remain a near-future prospect.

Availability and Requirements
BRIDE can be accessed with a web browser at <http://bride-db.eu/>. Data downloads are available in JSON format.

Conclusions
BRIDE is a knowledge broker for LDIR brain research and in particular systems radiobiology and omics. As is typical in databases, only positive cases are recorded i.e., genes that are known to be involved in LDIR response. Negative cases, i.e., genes not involved in LDIR response are much harder to capture as negative assertions. Further experimental measurements might augment the current count of 3174 records. Comparative genomics of the available content across other mammalian species is also a future prospect of great interest. Our implementation uses a simple serverless model, making all recorded data available to the community. BRIDE will hopefully find uses as a bioinformatics tool for the exposomics of the mammalian brain to LDIR.
Author Contributions: C.K., D.V. and C.A.O. designed, implemented and tested the database system. L.A. and C.A.O. coordinated and supervised the research and development. All authors read and approved the final manuscript.

Funding:
The early development of BRIDE has been wholly supported by the collaborative European project CEREBRAD (grant # 295552), within the 7th EU framework programme, Nuclear Fission and Radiation Protection. C.A.O. acknowledges support by the project Elixir-GR (grant # MIS 5002780), implemented under the action "Reinforcement of the Research and Innovation Infrastructure," funded by the Operational Program Competitiveness, Entrepreneurship, and Innovation (NSRF 2014-2020) and co-financed by Greece and the European Union (European Regional Development Fund).

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.

Data Availability Statement:
The data presented in this study are available in this article.