REPORT OF AN NSF-SUPPORTED WORKSHOP GLOBAL REGISTRY OF BIODIVERSITY REPOSITORIES: DESIGNING GRBIO VERSION 2 National Museum of Natural History, Smithsonian Institution, Washington, DC 27-28 April 2015 Summary The Smithsonian's National Museum of Natural History hosted an NSF-supported workshop to gather community input toward improvement of the Global Registry of Biodiversity Repositories (GRBio). Twenty-one workshop participants were invited as representatives of biodiversity collections and networks of collections, specimen databases, informatics standards and initiatives, and other online registries of collections. Participants felt that GRBio is valuable to researchers trying to locate repositories that house specimens cited in the literature. It could also provide publishers and others with a useful standard list of institutions and collections. It may have less impact in the future if the community implements machine-readable specimen identifiers on a global scale. Advice from participants addressed the registry's content, curation, user interface, and web services. The highest priorities were: creating web services that would exchange and synchronize records with other registries; enriching data records with summary information about an institution's or collection's digitized specimen records in other databases and the literature, and possibly pointers to those groups of records; implementing a protocol for displaying synonymies and homonymies among data elements; improving search and download capabilities; providing clearer online instructions to data submitters; and developing a community curation system that promotes participation without undue burden.

Background The Global Registry of Biodiversity Repositories (GRBio) is a web portal for a database of information about collections and the more than 7,000 institutions that contain them. The database also contains information on privately owned biodiversity collections and people associated with the institutions and collections. GRBio was created in 2013 by a developer working under contract for the Consortium for the Barcode of Life (CBOL)1 and the database was populated with records from three sources: 

www.biorepositories.org, an online database created by CBOL in 2007 with data from the institutional file maintained by the National Center for Biotechnology Information (NCBI). CBOL’s goal was to create an authoritative source for two terms in the Darwin Core data standard: institutionCodes2 and collectionCodes3.

1

CBOL is an initiative hosted by the Smithsonian’s National Museum of Natural History. Created in 2004 with support from the Alfred P. Sloan Foundation, CBOL’s mission is to develop and promote DNA barcoding as a global standard for species identification.

2

Darwin Core defines institutionCode as “The name (or acronym) in use by the institution having custody of the object(s) or information referred to in the record.”

3

collectionCode: “The name, acronym, coden, or initialism identifying the collection or data set from which the record was derived.”



Index Herbariorum (IH), a long-standing information resource for the botanical community. IH contains information on herbaria and their staff members, but unlike biorepositories.org and GRBio, IH does not distinguish between institutions and collections.



Biodiversity Collections Index (BCI), created around 2011 with records from IH and a database of insect and spider collections. BCI created more than 10,000 new collection records for the individual botanist collections noted in IH records and assigned LSIDs to all collections. BCI also established parent-child relationships between some collections but did not assign any to the rank of Institution. The database was taken offline when GRBio was released.

GRBio’s current structure and user interface were designed to accommodate the most important features of the three databases being merged, such as: 

The two-level collection-within-institution structure of biorepositories.org;



Incorporating staff member records and additional data fields from IH;



Retaining the LSIDs assigned to collections by BCI

The people responsible for IH (Barbara Thiers), BCI (Roger Hyam), and biorepositories.org (David Schindel) agreed that additional modification of GRBio would be beneficial based on feedback from the community. With this goal in mind, CBOL applied for and received support from NSF’s Advances in Biological Informatics Program for a community workshop.

Workshop Participants, Goals and Structure CBOL identified four major stakeholder groups with an interest in institution- and collection-level information. Twenty-one invited representatives of the following stakeholder groups participated in the workshop (see Appendix 1, Participant List): 

Other registries of institutions and collections (e.g., IH, the Global Biodiversity Information Facility [GBIF], World Federation of Culture Collections [WFCC]);



Collections, networks of collections (e.g., Canadensys, Consortium of European Taxonomic Facilities [CETAF]);



Collection databases (e.g., Specify, Arctos, VertNet, iDigBio); and



Biodiversity standards, initiatives, and platforms (e.g., Darwin Core, ZooBank, Biodiversity Information Serving Our Nation [BISON])

The goal of the workshop was to gather and prioritize suggestions from these stakeholder groups for improving GRBio in three areas: 

Data content, data fields, and controlled vocabularies used in them;



Functionality and the user interface; and



Interoperability with other information platforms and other web services.

The agenda for the workshop’s first day was devoted to short presentations by representatives of the four stakeholder groups followed by panel and whole-group discussions. The second day was devoted to synthesizing and prioritizing suggestions from the first day and developing strategies for promoting community participation and curation of the data in GRBio (see Appendix 2, Workshop Agenda, and presentations4). 4

PowerPoint presentations from the workshop are available at http://scicoll.org/events.html under Past Events, Washington, DC, April 2015

The Value and Impact of GRBio Enthusiasm for GRBio varied among the workshop participants due to differing perspectives and priorities. Some participants were more focused on access to physical specimens while others focused more on data records associated with those specimens. Participants who focused on physical specimens thought that GRBio would be valuable for raising the visibility and discoverability of collections whose specimens have not yet been digitized, especially smaller collections. From the specimen perspective, GRBio would also be useful for finding the current location of specimens starting with references in the published literature. These published references are often in the form of an institutionCode and catalogNumber, but specimens can move to different institutions and institutionCodes can change over time. These changes create a need for information on lineages of institutions, collections, and the names and codes by which they’re known. Participants whose primary focus was specimen data that having machine-readable ‘globally unique specimen identifiers (generally termed ‘GUIDs’) associated with all specimens would make traceability from literature citations to the specimens much more direct. In their view, replacing human-readable institutionCodes and catalogNumbers with GUIDs would render information about host institutions and collections relatively unimportant. Others felt similarly about collectionCodes: assigning GUIDs to collection records in GRBio would lead people to the specimens, so information about the parent institution would be unnecessary. Nevertheless, many participants saw the value of having human-readable institutionCodes and collectionCodes as a complement to machine-readable GUIDs. This was especially true for legacy publications and for print versions of journals going forward. There was universal agreement that institutions and collections should have GUIDs, either provided by them or minted by the registry (as BCI did with LSIDs and GRBio does with CoolURIs).

Recommendations: Data Content Integration of Registries. The presentations by other biorepository registries demonstrated the great variability of their content coverage. Some registries (e.g., Insect and Spider Collections of the World, Standard Symbolic Codes for Institutional Resource Collections in Herpetology & Ichthyology) consist primarily of institutionCodes, institutionNames, and address information. At the other extreme, registries like the Culture Collections Information Worldwide (CCINfo of the World Federation of Culture Collections) characterizes ~700 collections in great depth. The amount and type of information contained in GRBio places it in the middle of this range. Some workshop participants suggested that GRBio could attract more data contributors and users if it offered additional types of data about collections. Examples are dynamically updated connections to other platforms (e.g., GBIF, GenBank, Biodiversity Heritage Library, VertNet), specimen citations in the literature, or information streams on social media (e.g., Twitter, Facebook). Participants suggested conducting a comparison among the data fields used by GRBio, other leading registries, and the data fields recommended by the Natural Collections Descriptions Interest Group (NCD) of Biodiversity Information Standards (TDWG). CBOL could then propose a core set of data elements to TDWG as a practical standard for characterizing collections. Data Curation. GRBio is based on a community curation model which relies on representatives of institutions and collections or other knowledgeable specialists to create and update their data records in GRBio. In principle, anyone could edit any record in GRBio because there are no requirements for accreditation and no password protections, raising the risk of errors or vandalism. GRBio manages these risks by placing any new or edited record goes in a moderator queue for review by a CBOL staff member before it is published. This validation step is cursory and rarely involves contacting the institution or

collection to confirm the accuracy of a record. This led participants to discuss various alternative approaches to data validation but no consensus was reached. The alternatives, in roughly descending order of complexity, were: 

Create a member log-in system of vetted individuals who would be assigned different levels of authority to edit records (similar to ZooBank’s);



Create a member log-in system that includes only the primary contacts for institutions and collections (though this creates the challenge of first verifying who these people are);



‘Piggyback’ on an existing user system such as ORCID which would require a log-in step. The username would then be attached to the new or updated record which would create a disincentive against entering incorrect information or vandalizing data;



Appoint regional curators (like CETAF) or some other group of vetted individuals charged with creating and updating records based on information provided to them by nearby institutions and collections. This has the advantage of creating advocates for GRBio who would expand its content, but the disadvantage of concentrating the burden of curation rather than distributing it across the community;



Appoint regional moderators who would be knowledgeable enough to validate new and updated records. Regional moderators would have a lighter burden than regional curators;



Add a required field for the email address of the person submitting the data. This could be used to verify their authenticity and include it in the record.

Synonymies and homonymies. registries:

In his overview of GRBio, Schindel noted two added values over most

A. Homonymies. One of the obstacles to using the Darwin Core Triplet (the concatenation of institutionCodes, collectionCodes and catalogNumbers with a colon separator between them) as a unique specimen identifiers is the use of the same institutionCode by more than one institution. CBOL has done a variety of things to disambiguate these homonyms: 

Appending to denote Index Herbariorum records, thereby resolving many of the ambiguities between herbaria and museums using the same institutionCodes;



Identifying discontinued use of institutionCodes that reduced multiple uses; and



Encouraging institutions that were using multiple institutionCodes to abandon the use of ambiguous ones.

B. Synonymies. InstitutionCodes can change for a variety of reasons. An institution may close and the collections can become associated with a new institution, or an institution may re-brand itself with a new institutionCode. Some journals and databases change institutionCodes because word order is different in different languages which creates more synonyms. GRBio documents the connections among synonymous institutionCodes by: 

changing Status to Inactive in the old institution record and its associated collection records and including mention (and more recently hyperlinks) of the new institutionCode in the Description field, and



Including mention of the old institutionCode in the Description field of the new record.

Participants acknowledged the value of reducing ambiguities in cases of homonymy and synonymy but they suggested that these problems would always exist. Rather than trying to resolve them all, participants suggested that GRBio simply highlight them in some uniform way and try to establish a standard protocol for displaying lists of the homonyms or synonyms associated with each ambiguous

record. Records could also be flagged as ‘Clean’ if they are adequately populated and have no ambiguities due to synonymy or homonymy, versus ‘Dirty’ for incomplete and/or ambiguous records. Sub-collections. In addition to capturing data on collections within institutions, it could be valuable for GRBio to be able capture information on sub-collections (i.e., collections within collections). For example, an orphaned collection that is absorbed into another institution could be regarded as a separate collection or a sub-collection (e.g., Stanford University’s fish collection that went to the California Academy of Sciences). Institutions might also want to highlight some very special portion of a collection that is regionally or globally unique.

Recommendations: Functionality and user interface CBOL described the limitations encountered so far in using Drupal for database management. Specifically, it has been difficult to upload batches of new records and updates to existing records while the database is online. Participants suggested several strategies for using the Drupal interface management system in combination with a different database engine. Participants offered the following suggestions for improvements: 

Provide on-screen definitions of terms like ‘institution’ and ‘collection’ with guidance on how to enter information. This could reduce a lot of the variability in the way data have been entered. Providing this information on the data entry screens would be better than creating separate documents or FAQs;



Create separate screens for basic and advanced search functions, and add a keyword search capability and provide separate search fields for address, city, state and country in the advanced search, as an alternative for the combined location field in the basic search function;



Solve the problem of accents and other diacritical characters in data entries;



Provide drop-down menus with controlled vocabularies for states/provinces within countries;



Provide the ability to download single records or the sets of records produced by searches, in addition to the current option to download all records; and



Create an alerting service that would generate an email message to the Primary Contacts for collections and to Staff Members whenever changes are made to their records, as GRBio currently does for changes to institution records.

Recommendations: Interoperability and web services Synchronizing registries. Participants noted the many other active registries of biorepositories and that many of them would persist because of a strong sense of community ownership. Disciplines may prefer to retain their own collection registries and operate in parallel with GRBio, though they may be willing to synchronize/exchange records with GRBio. The workshop placed a high priority on developing web services that would synchronize records with other registries such as: 

Index Herbariorum



iDigBio



World Federation of Culture Collections



ISBER



American Society of Mammalogists



American Society of Ichthyologists and Herpetologists



Insect and Spider Collections of the World



Mammal collections in the Western Hemisphere



Atlas of Living Australia

Participants suggested other sources of records that could be added to GRBio in addition to registries. For example: 

Each installation of collection databases like Specify and Arctos could send a record to GRBio that could be incorporated if one didn’t already exist for that institution or collection;



Each dataset submitted to GBIF using the Integrated Publishing Tool could submit a record to GRBio about the collection(s) represented in the dataset

In many cases, the records GRBio records generated in this way may be sparsely populated. Nevertheless, having ‘stub’ records that can be updated is better than having an institution or collection absent from GRBio.

Recommendations: Building Community Participation Opinions varied widely among the workshop participants as to how the community would view GRBio’s value. The community would need to see clear benefits before their participation will increase substantially. Three factors emerged as most important: 1. Promoting the use of GRBio as an authoritative look-up resource that could be seen as an informal community certification, the way botanists view IH (though there is no formal evaluation done by IH or any specific certification). Creating a single certification standard for collections in industrialized and developing nations would be very difficult. Examples could be: - Representing the institutions with ‘Clean’ GRBio records to the Nagoya Protocol Clearinghouse as a group of registered, competent organizations; - Collection managers who use the ‘Clean’ list when deciding whether to honor loan requests from institutions they don’t know; - Holding institutions of type specimens registered in ZooBank could be required to be required to be registered in GRBio, and - Providing publishers with web services that would enable them to check whether the institutionCodes used in specimen references are registered and traceable; 2. Synchronization among different registries could be an important factor in promoting buy-in by enabling institutions to update their data in one place and have the update propagate to all registries; and 3. Enabling institutions and collections to submit information, or having GRBio assemble information automatically, that would improve their standing in the community. Examples of this information are: - Size of collections and their loan/visitor traffic; - Geographic and taxonomic coverage, and unique strengths; - Progress on digitization; - Amount of data contributed to GBIF or other aggregators; - Numbers of page views in GRBio; - Numbers of voucher specimens in GenBank; - Citations of voucher specimens and GenBank accession numbers in publications.

APPENDIX 1: List of Workshop Participants Participant

Institutional Affiliation

Representing

Shannon Asencio

Canadian Museum of Nature

Canadian Museum of Nature

Andrew Bentley

University of Kansas

Society for the Preservation of Natural History Collections (SPNHC) and Specify

Carol Butler

Smithsonian Institution

National Museum of Natural History

Dora Ann Lange Canhos

Reference Center on Environmental Information

CRIA

Ana Casino

Consortium of European Taxonomic Facilities

CETAF

Philippe Desmeth

Belgian Science Policy Office

World Federation of Culture Collections

Stinger Guala

US Geological Survey

Biodiversity Information Serving Our Nation (BISON)

Rob Guralnick

University of Florida

Map of Life, VertNet

Donald Hobern

Global Biodiversity Information Facility

GBIF, Atlas of Living Australia

Kim Labuschagne

South Africa National Zoological Garden

European, Middle Eastern & African Society for Biopreservation & Biobanking (ESBB)

Patricia Mergen

Royal Museum of Central Africa, Belgium

CETAF and SYNTHESYS

Francois Michonneau

University of Florida

iDigBio

Rod Page

University of Glasgow

GBIF

Cyndy Parr

USDA/National Agriculture Library

TDWG (Biodiversity Information Standards)

Richard Pyle

Bernice P. Bishop Museum

ZooBank and ICZN

Greg Riccardi

Florida State University

iDigBio

Susan Schafer

National Center for Biotechnology Information

GenBank/NCBI

David E. Schindel

Smithsonian Institution

Consortium for the Barcode of Life

David P. Shorthouse

University of Montréal

Canadensys

Barbara Thiers

New York Botanical Garden

Index Herbariorum

John Wieczorek

University of California, Berkeley

Darwin Core Standard, VertNet

APPENDIX 2: Workshop Agenda Monday, 27 April 2015, Smithsonian Castle Library 8:30am: Registration and continental breakfast 9:00am: Welcome, introductions, review of workshop goals and agenda 9:30am: Overview of the Global Registry of Biodiversity Repositories (David Schindel, Smithsonian) 10:00am: Session A: Current and potential relationships between GRBio and other online registries of collections: Content, Functionality and Interoperability (Chair: Susan Schafer, GenBank) 

10:00-10:40: Panelist Lightning talks (4 @ 10 minutes/10 slides each, maximum) - Index Herbariorum (Barbara Thiers) - GBIF (Donald Hobern) - World Federation of Culture Collections (Philippe Desmeth) - Atlas of Living Australia (Donald Hobern)  10:40-11:00: Q&A and discussion 11:00-11:30: Coffee break 11:30pm: Session B: Perspectives of Collections and Networks of Collections: Content, Functionality and Interoperability (Chair: Barbara Thiers, NY Botanical Garden) 

11:30-12:10: Panelist Lightning talks (4 @ 10 minutes/10 slides each, maximum) - National Museum of Natural History, Smithsonian Institution (Carol Butler) - Canadensys (David Shorthouse) - CETAF/SYNTHESYS (Ana Casino) - ISBER/ESBB (Kim Labuschange)  12:10-12:30: Q&A and discussion 12:30-1:30: Lunch 1:30pm: Session C: Perspectives of Collections Databases: Content, Functionality and Interoperability (Chair: Rob Guralnick, University of Florida) 

1:30-2:10: Panelist Lightning talks (4 @ 10 minutes/10 slides each, maximum) - iDigBio (Francois Michonneau) - Specify (Andy Bentley) - Arctos and VertNet (Laura Russell) - EMu (Brad Lickman)  2:10-2:45: Q&A and discussion 2:45-3:15: Coffee Break 3:15pm: Session D: Perspectives of other Biodiversity Informatics Standards, Initiatives and Platforms: Content, Functionality and Interoperability (Chair: Cyndy Parr, USDA National Agriculture Library) 

3:15-4:25 Panelist Lightning talks (7 @ 10 minutes/10 slides each, maximum) - GBIF (Rod Page) - iDigBio (Greg Riccardi) - ABCD (Patricia Mergen) - Darwin Core (John Wieczorek) - ZooBank and taxonomic publications (Rich Pyle) - Map of Life (Rob Guralnick) - BISON (Stinger Guala)  4:25-5:00: Q&A and discussion 5:00: Report of Rapporteur and discussion 5:30: Day 1 ends

Tuesday, 28 April 2015, Smithsonian Castle Library 8:30am: Continental breakfast 9:00am: Session E: What are the priorities for new specifications for Content, Functionality and Interoperability in GRBio Version 2.0? 

9:00-11:00am: Presentations by Day 1 Session Chairs and Group Discussion (Discussion Moderator: Andy Bentley, University of Kansas) - Session A. Other Registries (Susan Schafer) - Session B. Collections and Networks of Collections (Barbara Thiers) - Session C. Collection Databases (Rob Guralnick) - Session D. Biodiversity Informatics Platforms and Standards (Cyndy Parr) 11:00-11:30: Coffee Break 11:30: Session F: Improving Participation and Collaboration: (Discussion Moderator: David Schindel, Smithsonian)   

How can we promote data submission and curation by the community? How can we encourage other biodiversity informatics platforms to use GRBio data as a standard reference? How can GRBio contribute to community coherence and success in biodiversity informatics?

12:30: Lunch 1:30: Workshop ends