BBMRI-ERIC Directory: 515 Biobanks with Over 60 Million Biological Samples.

Biobanks[1] are well-organized repositories of biological material. They have become the fundamental resource for advancing medical research and constitute a major component of more generally understood bioresources. Yet they face a number of challenges to become more utilized on the national and global scale. These challenges range from fragmentation of data structure and sometimes even lack of availability of data,[2-4] lack of consistent quality management and traceability[5-8] to fragmentation of privacy protection regulations[9-13] and technical, organizational, and legal aspects of scalable secure storage and processing of privacy-sensitive big data.[14-16] To address the fragmentation and findability aspects, BBMRI-ERIC has released its Directory as a first IT service, providing aggregate information about the biobanks and bioresources. The Directory features a novel scalable distributed architecture, which enables updating data about changing resources in a long-term sustainable manner.

Inventory data about the bioresources, describing availability of various resource types such as biological material, data, expertise, and offered services, are the basis for any further interaction between the biobanks as resource/service providers and their users or collaborators. There have been various terms used for these types of services, including “catalogs” and “registries.” Inventory data cover various types of information that is not considered privacy sensitive and thus shareable in an open-access mode. The business model of a bioresource may impose access restrictions, however. From the users' perspective, it is important to achieve consistent or at least algorithmically harmonizable semantics of the information, so that it is possible to implement efficient search or filtering services.

There have been a number of attempts to improve the situation with availability and consistency of the inventory data in the past decade both internationally and nationally. Prominent international examples include P[3]G Observatory,[17] BBMRI Preparatory Phase Catalogue,[3] ISBER International Resource Locator,[18] Maelstrom Repository,[19] BBMRI-LPC catalogs,[20,21] or RD-CONNECT Catalogue[22,23] and the NIH/NCATS GRDR®[24] on rare diseases. Although being very valuable for helping to organize biobanking and bioresources in projects with limited life spans, these tools also demonstrate the key deficiency of such centrally built and managed systems: because of the lack of automated data updates, the information becomes sooner or later obsolete and thus of limited use for the users.

In contrast, distributed information systems are well known in computer infrastructures, such as cloud and grid computing systems,[25] where various architectures have been explored, ranging from client-server communication schemes[26,27] to peer-to-peer systems.[28-30] The biobanking community needs to learn from these endeavors and take a similar approach with (a) distributed architecture that allows for information flow from the original sources to the inventory services, (b) well-defined stable application programming interfaces (APIs) that allow for their implementation in the biobank information management systems, (c) clear component-based architecture that allows for simple implementation of relevant data extraction and harmonization components as close to the original information sources as possible to include in-depth knowledge of the data.

BBMRI-ERIC Directory

BBMRI-ERIC, the Biobanking and BioMolecular Resources Research Infrastructure-European Research Infrastructure Consortium, is a new form of legal organization for biobanking in Europe. BBMRI-ERIC has started to develop the BBMRI-ERIC Directory as its first information technology tool. Directory 1.0 was released in July 2015 with basic support for biobanks, and Directory 2.0 was released in December 2015, supporting biobanks, collections, and biobank networks. The BBMRI-ERIC Directory has been designed with the following two primary use cases in mind: biomedical and bioinformatics researchers seeking retrieval of samples/data, or for collecting/hosting services for their samples/data; biobank operators needing to identify similar biobanks (experience sharing, collaboration, etc.) and to promote their visibility. One can also imagine other use cases such as research participants (donors/patients) and their organizations interested in determining where their samples might be located, the purposes they are being used, and for funding and governance bodies looking into the extent and use statistics of funded infrastructures.

The first development step was designing an extensible data model, which covers all three key components of biobanks: (a) biological material and associated physical storage facilities, (b) data and associated data storage facilities, and (c) expertise of the biobankers. The core of the data model for Directory 2.0 relies on MIABIS 2.0,[31] which is the evolution of the previously published MIABIS model.[32] The Directory's data model includes biobanks as institutional units hosting collections of samples and data, as well as biobank networks used to further aggregate biobanks or their collections. Our current data model is highly aggregated and serves to primarily identify candidate biobanks that might have samples for the given purpose or biobanks that provide relevant services.

From the architectural perspective, the Directory is a distributed system using multilayer architecture as shown in Figure 1. Each layer uses clearly defined machine-readable APIs (LDAP, REST/JSON) and data formats, which enable automated propagation of updates, allows for building purpose-focused user interfaces, as well as integrates into larger automated workflows (e.g., currently developed BBMRI-ERIC Negotiator to facilitate access negotiation). As for Directory 2.0, there are two web-based user interfaces implemented: the main BBMRI-ERIC Directory interface[33] and the BBMRI.nl interface,[34] integrating BBMRI-ERIC data using the Molgenis platform.[35]

Distributed and modular BBMRI-ERIC Directory architecture. It typically comprises online or offline data flow from biobanks and other bioresources (not shown for sake of simplicity) → aggregating nodes (e.g., BBMRI-ERIC national nodes) → central BBMRI-ERIC Directory → user interfaces. Color images available online at www.liebertpub.com/bio

The current Directory 2.0 includes 515 biobanks and standalone collections, with an estimated number of samples exceeding 60,000,000. This covers 136 clinical or disease-specific biobanks and 189 population biobanks, based on the classification proposed in an article[3] from the BBMRI Preparatory Phase. This is a conservative estimate based on the 10n order of magnitude attribute, which is mandatory for each collection in Directory 2.0, compared with optional exact size. For the largest biobanks, the estimate has been adjusted based on direct communication to avoid substantial bias. We consider these estimates sufficient, as exact counting would require consensus on sample and aliquot definition (expected to be clarified by ISO TC 276 between 2017 and 2018).[*]

It should also be noted that the access to the samples/data is controlled by the biobanks, which means biobanks may or may not allow access depending on the types of requests received by them. Based on Directory 2.0 data, ∼23% of biobanks provide access to samples/data based on a fee structure and ∼28% based on joint projects. Approximately 60% do not publish this information and need to be contacted directly to receive information on access conditions.

Future Work

There are two basic directions to improve the inventory services, and particularly the BBMRI-ERIC Directory. The first direction is to improve specificity in the responses. Although the biobanks are already urged to publish data that are as accurate as possible, many issues will only be resolved once we help biobanks fully implement online interfaces to their primary information systems. A particular example is the list of available diagnoses, which is among the most searched for parameters.[36] Some biobanks do not have this information themselves and must either retrieve it using targeted questions to hospital information systems or must resort to consulting external registries.

The second direction for the future extensions is improving coverage of various aspects of the biobanks, such as availability of quality management systems, advertising additional services such as sample/data hosting, or providing semantic translation support for data that comes in different coding or semantics from different sources. The Directory service can also be used to map various types of identifiers and to publish persistent identifiers for sample sets and datasets once they are used for publishing, hence further supporting efforts toward reproducible biomedical research. These extensions are expected in Directory 3.0 (to be released in 2016) and onward. BBMRI-ERIC also works on extending geographical coverage of the Directory by merging with the validated RD-CONNECT Catalogue data during 2016 (e.g., data from the United States and Australia).

Last but not least, the communication of many users with many biobanks at the same time is not efficient, and tools for simplifying such communication are needed. BBMRI-ERIC will address these issues using the Negotiator tool integrated with the Directory, intended for cumulative communication between a user and multiple biobanks at the same time.

Because of its potential global impact, the Directory has been proposed as a tool for organizing bioresources' inventory information as a part of the BBMRI-ERIC application for the G7 Group of Senior Officials on Global Research Infrastructures.