FAIRness in Biomedical Data Discovery

Alina Trifan and Jos

e Lu

ıs Oliveira

IEETA/DETI, University of Aveiro, Portugal

Keywords:

Biomedical Data Discovery, FAIR Guidelines, Data Discovery Platforms, FAIR Metrics.

Abstract:

The FAIR Guiding Principles are a recent, yet powerful set of recommendations for turning data Findable,

Accessible, Interoperable and Reusable. They were designed with the purpose of improving data quality and

reusability. Over the last couple of years they have been adopted more and more by both data owners and

funders as key data management approaches. Despite their increasing popularity and endorsement by multiple

research initiatives from some of the most diverse areas, there are still only a few examples on how these

principles have been translated into practice. In this work we propose an open evaluation of their adoption

by biomedical data discovery platforms. We ﬁrst overview current biomedical data discovery platforms that

introduce the FAIR guiding principles as requirements of their functioning. We then employ the more recent

FAIR metrics for evaluating the degree to which these biomedical data discovery platforms follow the FAIR

principles. Moreover, we assess their impact on enabling data interoperability and secondary reuse.

1 INTRODUCTION

The FAIR guiding principles - FAIR stands for Find-

able, Accessible, Interoperable and Reusable - were

proposed with the ultimate goal of reusing valuable

research objects (Wilkinson et al., 2016). They rep-

resent a set of guidelines for turning data more mean-

ingful and reusable and they emphasize the need of

making data discoverable and interoperable by ma-

chines. These principles do not provide strict rules

or standards to comply with, but rather focus on con-

ventions that enable data interoperability, stewardship

and compliance against data and metadata standards,

policies and practices.

They are not standards to be rigorously followed,

but rather permissive guidelines. The principles are

aspirational, in that they do not strictly deﬁne how to

achieve a state of FAIRness. Depending on the needs

or constraints of different research communities, they

can be open to interpretation. Independently of this

openness, they were designed to assist the interaction

between those who want to use community resources

and those who provide them. When followed, they

are beneﬁcial for both data owners and users that seek

access to the data. These principles have rapidly been

adopted by publishers, funders, and pan-disciplinary

infrastructure programmes as key data management

issues to be taken into consideration. This can be ex-

plained as data management closely relates to inter-

operability and reproducibility (Jansen et al., 2017).

Generic and research-speciﬁc initiatives, such as

the European Open Science Cloud

, the European

Elixir infrastructure

and the USA National Institutes

of Health’s Big Data to Knowledge Initiative

are

some of the current initiatives endorsing the FAIR

principles and commiting to provide FAIR ecosys-

tems across multi-disciplinary research areas. More-

over, the European Commission has recently made

available a set of recommendations and demands for

open data research that are explicitly written in the

context of FAIR data

With regard to biomedical data sources, data inter-

operability and reusability has been a hot topic over

the last decade, strongly correlated with the evolu-

tion of the so called Big Data in Healthcare. De-

spite the incremental increase of the use and storage

of electronic health records, the biomedical commu-

nity still tends to use these data in isolation. Unfortu-

nately more than 80% of the datasets in current prac-

tice are effectively unavailable for reuse (Mons et al.,

2017). This is just one of the factors behind the repro-

ducibility crisis that is manifesting in the biomedical

http://eoscpilot.org

http://www.elixir-europe.org

http://commonfound.nih.gov/bd2k/

http://ec.europa.eu/research/participants/data/ref/

h2020/grants manual/hi/oa pilot/h2020-hi-oa-data-

mgt en.pdf

Trifan, A. and Oliveira, J.

FAIRness in Biomedical Data Discovery.

DOI: 10.5220/0007576401590166

In Proceedings of the 12th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2019), pages 159-166

ISBN: 978-989-758-353-7

159

arena (Prinz et al., 2011). Apart from data still be-

ing gathered in silos unavailable outside of the owning

institution or country, data privacy concerns and un-

clear data management approaches are critical barri-

ers for sharing and reusing data. The FAIR principles

have been enabling the global debate about better data

stewardship in data-driven and open science, and they

have triggered funding bodies to discuss their applica-

tion to biomedical systems. A wide adoption of these

principles by the data sources and systems that handle

biomedical data has the ability to leverage this repro-

ducibility crisis, by ensuring interoperability among

heterogeneous data sources.

In this paper we propose an overview of the adop-

tion of these principles by biomedical data discov-

ery platforms, which are considered important en-

ablers of secondary research. In the process of reusing

biomedical data for secondary research, discovery

platforms play an important part as they provide sup-

port for identifying data sources that can answer a

given translational research question. We review cur-

rent biomedical data platforms in order to under-

stand their level of FAIRness. While some of these

platforms already performed self-assesments of their

methodologies for following the FAIR principles, our

exhaustive literature search revealed only a handful

of such assessments. Otherwise the FAIR principles

are identiﬁed as system requirements of several dis-

covery platforms, without a deep evaluation of their

adoption. We therefore chose three such platforms

and identify the means through which they answer to

the FAIR guidelines.

This paper is structured in 5 more sections. A de-

tailed presentation of the FAIR principles is covered

in Section 2, followed by an overview of biomedical

data discovery platforms in Section 3. The adoption

of the FAIR principles by these platforms is analyzed

in Section 4. We discuss the importance of this adop-

tion for the biomedical research community in Sec-

tion 5 and we draw the ﬁnal remarks in Section 6.

2 FAIR GUIDING PRINCIPLES

The FAIR principles were intended as a set of guide-

lines to be followed in order to enhance the reusabil-

ity of any type of data. They put speciﬁc emphasis

on enhancing the ability of machines to automatically

ﬁnd and (re)use the data, in addition to supporting

its (re)use by individuals. The goal is that, through

the pursuit of these principles, the quality of a data

source becomes a function of its ability to be accu-

rately found and reused. Although they are currently

not a strict requirement, nor a standard in biomedi-

cal data handling systems, these principles maximize

their added-value, by acting as a guidebook for safe-

guarding transparency, reproducibility, and reusabil-

ity.

The FAIR principles as initially proposed

by (Wilkinson et al., 2016) are detailed in Table 1.

In a nutshell, if a data source is intended to be FAIR,

sufﬁcient metadata must be provided to automat-

ically identify its structure, provenance, licensing

and potential uses, without having the need to use

specialized tools. Moreover, any access protocols

should be declared where they do or do not exist. The

use of vocabularies and standard ontologies further

beneﬁt to the degree of FAIRness of a data set.

The way these principles should manifest in re-

ality was largely open to interpretation and more

recently some of the original authors revisited the

principles, in an attempt to clarify what FAIRness

is (Mons et al., 2017). They addressed the principles

as a community-acceptable set of rules of engagement

and a common denominator between those who want

to use a community’s resources and those who pro-

vide them. An important clariﬁcation was that FAIR

is not a standard and it is not equal to open. The ini-

tial release of the FAIR principles were somehow mis-

leading in the sense that accessibility was associated

with open access. Instead, in the recent extended ex-

planation of what these principles really mean, the A

in FAIR was redeﬁned as “Accessible under well de-

ﬁned conditions”. This means that data do not have to

be open, but the data access protocol should be open

and clearly deﬁned. In fact, data should be “as open

as possible, as closed as needed”.

The recognition that computers must be capable

of accessing a data object autonomously was the core

to the FAIR principles since the beginning. The re-

cent re-interpretation of these principles maintains

their focus on the importance of data being accesi-

ble to autonomous machines and further clariﬁes on

the possible degrees of FAIRness. While there is no

such notion as unFAIR, the authors discuss the dif-

ferent levels of FAIRness that can be achieved. As

such, the addition of rich, FAIR metadata is the most

important step towards becoming maximally FAIR.

When data objects themselves can be made FAIR and

open for reuse, the highest degree of FAIRness can

be achieved. When all of these are linked with other

FAIR data, the Internet of FAIR data is reached. Ulti-

mately, when a large number of applications and ser-

vices can link and process FAIR data, the Internet of

FAIR Data and Services is attained.

HEALTHINF 2019 - 12th International Conference on Health Informatics

160

Table 1: The FAIR Guiding Principles as originally proposed in (Wilkinson et al., 2016).

Findable F1. (meta)data are assigned a globally unique and persistent identiﬁer.

F2. data are described with rich metadata (deﬁned by R1 below).

F3. metadata clearly and explicitly include the identiﬁer of the data it describes.

F4. (meta)data are registered or indexed in a searchable resource.

Accessible

A1. (meta)data are retrievable by their identiﬁer using a standardized

communications protocol.

A1.1 the protocol is open, free, and universally implementable.

A1.2 the protocol allows for an authentication and authorization procedure,

where necessary.

A2. metadata are accessible, even when the data are no longer available.

Interopera-

ble

I1. (meta)data use a formal, accessible, shared, and broadly applicable language

for knowledge representation.

I2.(meta)data use vocabularies that follow FAIR principles.

I3.(meta)data include qualiﬁed references to other (meta)data.

Reusable

R1. (meta)data are richly described with a plurality of accurate and relevant

attributes.

R1.1 (meta)data are released with a clear and accessible data usage license.

R1.2 (meta)data are associated with detailed provenance.

R1.3 (meta)data meet domain-relevant community standards.

3 BIOMEDICAL DISCOVERY

PLATFORMS

The integration and reuse of huge amounts of biomed-

ical data currently available in digital format has the

ability to impact clinical decisions, pharmaceutical

discoveries, disease monitoring and the way popula-

tion healthcare is provided globally. Storing data for

future reuse and reference has been a critical factor

in the success of modern biomedical sciences (Raz-

ick et al., 2014). In order for data to be reused, ﬁrst

it has to be discovered. Finding a dataset for a study

can be burdensome due to the need to search individ-

ual repositories, read numerous publications and ulti-

mately contact data owners or publication authors on

an individual basis. Recent research shows that the

time spent by researchers in searching for and identi-

fying multiple useful data sources can take up to 80%

of their time dedicated to the project or research ques-

tion itself (Press, 2016).

Biomedical data exists in multiple scales, from

molecular to patient data. Health systems, genetics

and genomics, population and public health are all

areas that may beneﬁt from big data integration and

its associated technologies (Martin-Sanchez and Ver-

spoor, 2014). The secondary reuse of citizens’ health

data and investigation of the real evidence of ther-

apeutics may lead to the achievement of personal-

ized, predictive and preventive medicine (Phan et al.,

2012). However, in order for researchers to be able to

reuse data and conduct integrative studies, they ﬁrst

have to ﬁnd the right data for their research. Data dis-

covery platforms are one-stop shops that enable clini-

cal researchers to identify datasets of interest without

having to perform individual, extensive searches over

distributed, heterogeneous health centers.

There are currently many data discovery plat-

forms, developed either as warehouses or simply ag-

gregators of metadata that link to the original data

sources. A warehouse platform, the Vanderbilt ap-

proach (Danciu et al., 2014) contains both fully de-

identiﬁed research data and fully identiﬁed research

that is made available taking into consideration access

protocols and governance rules. A cataloguing toolkit

is proposed by Maelstrom Research, built upon two

main components: a metadata model and a suite of

open-source software applications (Bergeron et al.,

2018). When combined, the model and software

support implementation of study and variable cata-

logues and provide a powerful search engine to facil-

itate data discovery. Disease oriented platforms, such

as The Ontario Brain Institute’s (Brain-CODE) (Vac-

carino et al., 2018) are designed with a very explicit,

yet not limited, purpose of supporting researchers

in better understanding a speciﬁc disease. Brain-

CODE addresses the high dimensionality of clinical,

neuroimaging and molecular data related with vari-

ous brain conditions. The platform makes available

integrated datasets that can be queried and linked

to provincial, national and international databases.

Similarly, the breast cancer (B-CAN) platform (Wen

et al., 2017) was designed as a private cancer data

center that enables the discovery of cancer-related

FAIRness in Biomedical Data Discovery

161

data and drives research collaborations aimed at bet-

ter understanding this disease. Still in the spectrum of

cancer discovery, the Project Data Sphere was built

to voluntarily share, integrate, and analyze histori-

cal cancer clinical trial data sets with the ﬁnal goal

of advancing cancer research (Green et al., 2015).

In the rare disease spectrum, RD-Connect (Gainotti

et al., 2018) links genomic data with patient reg-

istries, biobanks, and clinical bioinformatics tools in

an attempt to provide a FAIR rare disease complete

ecosystem.

Among most established initiatives, Cafe Var-

iome (Lancaster et al., 2015) provides a general-

purpose, web-based, data discovery tool that can be

quickly installed by any genotype–phenotype data

owner and turn data discoverable. MONTRA (Silva

et al., 2018), another full-ﬂedged open-source dis-

covery solution, is a rapid-application development

framework designed to facilitate the integration and

discovery of heterogeneous objects. Both solutions

rely on a catalogue for data discovery and include ex-

tensive search functionalities and query capabilities.

Linked Data is also explored in discovery plat-

forms, such as YummyData (Yamamoto et al., 2018)

which was designed to improve the ﬁndability and

reusability of life science datasets provided as Linked

Data. It consists of two components, one that period-

ically polls a curated list of SPARQL endpoints and a

second one that monitors them and presents the infor-

mation measured. Similarly, the Open PHACTS Dis-

covery Platform (Groth et al., 2014) leverages Linked

Data to provide integrated access to pharmacology

databases. Still in the spectrum of Linked Data,

BioSharing is a manually curated searchable portal of

three linked registries (McQuilton et al., 2016) that

cover standards, databases and data policies in the life

sciences.

All these platforms address data discovery from

different perspectives, integrating or linking to differ-

ent types of biomedical data. Another aspect that they

share is that they identify the FAIR principles as re-

quirements of their architectures, as well as enablers

of data discovery. Although the high majority of these

platforms emphasize the importance of providing a

way for machines to discover and access the data sets,

they are heterogeneous in the way they address the

FAIR guidelines. For this evaluation, we have chosen

three of the previously overviewed data discovey plat-

forms for understanding their approaches in following

the FAIR guiding principles. We ﬁrst overview the

scope and methods of these platforms and we present

in a narrative form their partial or total compliance

with the FAIR principles.

4 ADOPTION OF THE FAIR

PRINCIPLES

In lifesciences, initiatives such as GOFAIR

make

use of infrastructures that already exist in European

countries to create a federated approach for turning

the FAIR principles a working standard in science.

Dataverse (Magazine, 2011), for instance, is an open-

source data repository software designed to support

public community or institutional research reposito-

ries. Another example is FAIRDOM

, a web plat-

form built for collecting, managing, storing, and pub-

lishing data, models, and operating procedures. Both

solutions follow the FAIR guiding principles in an

attempt to improve research management practices.

Open PHACTS

, a data integration platform for drug

discovery, UniProt (Pundir et al., 2017), an online re-

source for protein sequence and annotation data and

the EMIF Catalogue (Trifan and Oliveira, 2018), are

some of the few FAIR self-assessed data discovery

and integration platforms.

Among the three platforms we chose for this as-

sessment, the Maelstrom Research cataloguing toolkit

presented by (Bergeron et al., 2018) is built upon two

main components: a metadata model and a suite of

open-source software applications. The model sets

out speciﬁc ﬁelds to describe study proﬁles, charac-

teristics of the subpopulations of participants, timing

and design of data collection events and variables col-

lected at each data collection event. The model and

software support implementation of study and vari-

able catalogues and provide a powerful search en-

gine to facilitate data discovery. Developed as an

open source and generic tool to be used by a broad

range of initiatives, the Maelstrom Research cata-

loguing toolkit serves several national and interna-

tional initiatives. The FAIR principles have been

identiﬁed from early on as a requirement of its ar-

chitecture. With respect to Findability, each dataset

is complemented by rich metadata. To ensure qual-

ity and standardization of the metadata documented

across networks, standard operating procedures were

implemented. In what concerns Accessibility, when

completed, study and variable-speciﬁc metadata are

made publicly available on the Maelstrom Research

website. Using information found in peer-reviewed

journals or on institutional websites, the study out-

line is documented and validated by study investiga-

tors. Thus, the linkage with other FAIR metadata is

achieved. Where possible, data dictionaries or code-

http://go-fair.org

https://fair-dom.org/about-fairdom/

http://www.openphactsfoundation.org/

HEALTHINF 2019 - 12th International Conference on Health Informatics

162

books are obtained, which contributes to the data in-

teroperability.

Many life science datasets are nowadays repre-

sented via Linked Data technologies in a common

format (the Resource Description Framework). This

makes them accessible via standard APIs (SPARQL

endpoints), which can be understood as one of the

FAIR requirements. While this is an important step

toward developing an interoperable bioinformatics

data landscape it also creates a new set of obsta-

cles as it is often difﬁcult for researchers to ﬁnd the

datasets they need. YummyData provides researchers

the ability to discover and assess datasets from differ-

ent providers (Yamamoto et al., 2018). This assess-

ment can be done in terms of metrics such as service

stability or metadata richness. YummyData consists

of two components: one that periodically polls a cu-

rated list of SPARQL endpoints monitoring the states

of their Linked Data implementations and content and

another one that presents the information measured

for the endpoints and provides a forum for discus-

sion and feedback. It was designed with the purpose

to improve the ﬁndability and reusability of life sci-

ence datasets provided as Linked Data and to foster its

adoption. Apart from making data available to soft-

ware agents via an API, the adoption of Linked Data

principles has the potential to make data FAIR.

BioSharing is a manually curated searchable por-

tal of three linked registries (McQuilton et al., 2016).

These resources cover standards, databases and data

policies in the life sciences broadly encompassing the

biological environmental and biomedical sciences.

The manifest of the initiative is that BioSharing

makes these resources ﬁndable and accessible - the

core of the FAIR principle. Every record is designed

to be interlinked providing a detailed description not

only on the resource itself but also on its relations

with other life science infrastructures. BioSharing is

working with an increasing number of journals and

other registries and its focus is to ensures that data

standards, biological databases and data policies are

registered, informative and discoverable. Thus, it is

considered a pivotal resource for the implementation

of the ELIXIR-supported FAIR principles.

4.1 FAIR Metrics

Along with the narrative analysis of their FAIR ap-

proaches, we propose an assessment following the

FAIR metrics recently proposed by some of the orig-

inal authors of the FAIR guiding principles (Ta-

ble 2). The increasing ambiguity behind the ini-

tially published principles, along with the need of

data providers and regulatory bodies to evaluate their

translation into practice led to the establishment of

the FAIR metrics group

, with the purpose of deﬁn-

ing universal measures of data FAIRness. Neverthe-

less, these universal metrics can be complemented by

resource-speciﬁc ones that can reﬂect the expectations

of one or multiple communities.

The second part of our assessment follows the pre-

viously identiﬁed FAIRness metrics, applied to each

of the 13 items of the FAIR guiding principles. For

each of the principles, we outline next the questions

that we tried to answer in the evaluation and the name

of the metric, within brackets. The following infor-

mation is a summary of the FAIR metrics description

proposed by some of the original authors of the guid-

ing principles

• F1 (Identiﬁer uniqueness) Whether there is a

scheme to uniquely identify the digital resource.

• F1 (Identiﬁer persistence) Whether there is a pol-

icy that describes what the provider will do in the

event an identiﬁer scheme becomes deprecated.

• F2 (Machine-readability of metadata) The avail-

ability of machine-readable metadata that de-

scribes a digital resource.

• F3 (Resource identiﬁer in metadata) Whether the

metadata document contains the globally unique

and persistent identiﬁer for the digital resource.

• F4 (Indexed in a searchable resource) The degree

to which the digital resource can be found using

web-based search engines.

• A1.1 (Access Protocol) The nature and use limi-

tations of the access protocol.

• A1.2 (Access authorization) Speciﬁcation of a

protocol to access restricted content.

• A2 (Metadata longevity) The existence of meta-

data even in the absence/removal of data.

• I1 (Use a knowledge representation language) The

use of a formal, accessible, shared, and broadly

applicable language for knowledge representa-

tion.

• I2 (Use FAIR Vocabularies) The metadata val-

ues and qualiﬁed relations should themselves be

FAIR, for example, terms from open, community-

accepted vocabularies published in an appropriate

knowledge-exchange format.

• I3 (Use qualiﬁed references) Relationships within

(meta)data, and between local and third-party

data, have explicit and ’useful’ semantic meaning.

http://fairmetrics.org

https://github.com/FAIRMetrics/Metrics/blob/master/

ALL.pdf

FAIRness in Biomedical Data Discovery

163

Table 2: The template for creating FAIR Metrics retrieved from https://github.com/FAIRMetrics.

FIELD DESCRIPTION

Metric Identiﬁer

FAIR Metrics should, themselves, be FAIR objects, and thus

should have globally unique identiﬁers.

Metric Name A human-readable name for the metric.

To which principle does it

apply

Metrics should address only one sub-principle, since each FAIR

principle is particular to one feature of a digital resource; metrics

that address multiple principles are likely to be measuring multiple

features, and those should be separated whenever possible.

What is being measured

A precise description of the aspect of that digital resource that is

going to be evaluated.

Why should we measure it Describe why it is relevant to measure this aspect.

What must be provided What information is required to make this measurement?

How do we measure it In what way will that information be evaluated?

What is a valid result What outcome represents “success” versus “failure”?

For which digital

resource(s) is this relevant

If possible, a metric should apply to all digital resources; however,

some metrics may be applicable only to a subset. In this case, it is

necessary to specify the range of resources to which the metric is

reasonably applicable.

Example of their

application across types of

digital resource

Whenever possible, provide an existing example of success, and an

example of failure.

• R1.1 (Accessible Usage License) The existence of

a license document, for both (independently) the

data and its associated metadata, and the ability to

retrieve those documents.

• R1.2 (Detailed Provenance) That there is prove-

nance information associated with the data, cov-

ering at least two primary types of provenance in-

formation: who/what/when produced the data (i.e.

for citation) and why/how was the data produced

(i.e. to understand context and relevance of the

data).

• R1.3 (Meets Community Standards) Certiﬁcation,

from a recognized body, of the resource meeting

community standards.

This evaluation allowed us to identify the FAIR re-

quirements already satisﬁed and the ones that are not

undressed, or unclear. Our ﬁndings show a high level

of FAIRness achieved by the three platforms, mainly

favored by the rich metadata with which each of these

platform complement the actual data sources. In all

cases the metadata can be accessed both by humans

and machines through a unique and persistent iden-

tiﬁer, mostly in the form of an URI. Moreover, the

use of FAIR standards and vocabularies contributes

to their degree of FAIRness. This is complemented

in two of the platforms by the ability to link to other

FAIR metadata, which speaks for the data interoper-

ability and reusability. Still related to reusability, the

use of Linked Data by two of the platforms is one of

its strong enablers. Last but not least, all of the plat-

forms support machine discoverability and access, by

providing dedicated APIs. The main unclear aspect

was the access protocol, which was not trivial to iden-

tify. Another weak point was the lack of quantiﬁable

certiﬁcation that the resources meet community stan-

dards. We present our summarized assessment in Ta-

ble 3.

5 DISCUSSION

Researchers need tools and support to manage, search

and reuse data as part of their research work. In the

biomedical area, data discovery platforms, either in

the shape of data warehouses or metadata integrators

that link to original data silos support the researcher

in the process of ﬁnding the right data for a given re-

search topic. However, ﬁnding the right data is not

sufﬁcient for conducting a study. Data should be not

only qualitative and accessible under clear and well-

deﬁned protocols, but it should also be interoperable

and reusable in order to maximize the research out-

comes. The FAIR guiding principles are recommen-

dations on the steps to follow in order to increase the

meaningfulness and impact of data and are strongly

related to data management. FAIR compliant biomed-

ical data discovery platforms have the ability to sup-

port biomedical researchers throughout all the steps

from ﬁnding the right data source to reusing it for

HEALTHINF 2019 - 12th International Conference on Health Informatics

164

Table 3: Assessment of the FAIRness of each of the three discovery platforms based on the FAIRness metrics. X represents a

satisﬁed requirement and - means that no proof to support the requirement was found.

Platform

F1 F2 F3 F4 A1.1 A1.2 A2 I1 I2 I3 R1.1 R1.2 R1.3

Maelstrom catalogue X X X X X – – X X X – X –

YummyData X X X X X X – X X – X X –

Biosharing X X X X X X – X X X X X –

secondary research. This can ultimately lead to bet-

ter health and healthcare outcomes. Ultimately, these

principles give an important contribution to the repro-

ducibility of research.

The FAIR guiding principles have been widely

endorsed by publishers, funders, data owners and

innovation networks across multiple research areas.

Up until recently, they did not strictly deﬁne how

to achieve a state of FAIRness and this ambiguity

led to some qualitatively different self-assessments of

FAIRness. A new template for evaluating the FAIR-

ness of a data set or a data handling system, recently

proposed by some of the original authors of the prin-

ciples, offers a benchmark for a standardized evalua-

tion of such self-assessments. In this paper we have

applied them to three different biomedical data dis-

covery platforms in order to estimate their FAIRness.

Moreover, we sought to understand the impact that

the adoption of these guidelines has in the quality of

the output produced by these platforms and to what

degree ensuring data reusability and interoperability

turns data more prone to be reused for secondary re-

search.

This analysis revealed that the adoption of

the FAIR principles is an ongoing process within

the biomedical community. However, the FAIR-

compliance of a resource or system can be distinct

from its impact. The platforms discussed exposed a

high level of FAIRness and an increased concern for

enabling data discovery by machines. While FAIR is

not equal to Linked Data, Semantic Web technologies

along with formal ontologies fulﬁll the FAIR require-

ments and can contribute to the FAIRness of a discov-

ery platform.

With digital patient data increasing at an expo-

nential rate and having understood the importance of

reusing these data for secondary research purposes, it

is highly important to ensure its interoperability and

reusability. The assessment of data FAIRness is a

key element for providing a common ground for data

quality to be understood by both data owners and data

users. If up until recently the open interpretation of

the FAIR guidinig principles could lead to assessment

biases, the recently published FAIR metrics support

more than ever the implementation of the common

ground. For this, the biomedical research community

should continue to challenge and reﬁne their imple-

mentation choices in order to achieve a desirable In-

ternet of FAIR Data and Services.

6 CONCLUSIONS

The FAIR principles demand well-deﬁned qualities

and properties from data resources but at the same

time they allow a great deal of freedom with respect

to how they should be implemented. In this work

we evaluated the approaches followed by three differ-

ent biomedical data discovery platforms in providing

FAIR data and services. This evaluation was strictly

done based on the analysis of the scientiﬁc publica-

tions describing these platforms. As future work, we

intend to extend this assessment by exploring these

platforms hands-on, in an attempt to address spe-

ciﬁc driving medical questions. Nevertheless, these

fresh examples highlighted the increasing impact of

the FAIR principles among the biomedical research

community. Moreover, by acting in accordance with

the FAIR metrics we, as a community, can reach an

agreed basis for the assessment of data quality.

REFERENCES

Bergeron, J., Doiron, D., Marcon, Y., Ferretti, V., and

Fortier, I. (2018). Fostering population-based cohort

data discovery: The Maelstrom Research cataloguing

toolkit. PloS one, 13(7):e0200926.

Danciu, I., Cowan, J. D., Basford, M., Wang, X., Saip, A.,

Osgood, S., Shirey-Rice, J., Kirby, J., and Harris, P. A.

(2014). Secondary use of clinical data: the Vanderbilt

approach. Journal of biomedical informatics, 52:28–

35.

Gainotti, S., Torreri, P., Wang, C. M., Reihs, R., Mueller,

H., Heslop, E., Roos, M., Badowska, D. M., Paulis, F.,

Kodra, Y., et al. (2018). The RD-Connect Registry &

Biobank Finder: a tool for sharing aggregated data and

metadata among rare disease researchers. European

Journal of Human Genetics, 26(5):631.

Green, A. K., Reeder-Hayes, K. E., Corty, R. W., Basch, E.,

Milowsky, M. I., Dusetzina, S. B., Bennett, A. V., and

Wood, W. A. (2015). The project data sphere initia-

FAIRness in Biomedical Data Discovery

165

tive: accelerating cancer research by sharing data. The

oncologist, 20(5):464–e20.

Groth, P., Loizou, A., Gray, A. J., Goble, C., Harland, L.,

and Pettifer, S. (2014). Api-centric linked data inte-

gration: the open PHACTS discovery platform case

study. Web Semantics: Science, Services and Agents

on the World Wide Web, 29:12–18.

Jansen, C., Beier, M., Witt, M., Frey, S., and Krefting, D.

(2017). Towards reproducible research in a biomed-

ical collaboration platform following the FAIR guid-

ing principles. In Companion Proceedings of the10th

International Conference on Utility and Cloud Com-

puting, pages 3–8. ACM.

Lancaster, O., Beck, T., Atlan, D., Swertz, M., Thangavelu,

D., Veal, C., Dalgleish, R., and Brookes, A. J. (2015).

Cafe Variome: General-purpose software for making

genotype–phenotype data discoverable in restricted or

open access contexts. Human mutation, 36(10):957–

964.

Magazine, D.-L. (2011). The dataverse network

: an

open-source application for sharing, discovering and

preserving data. D-lib Magazine, 17(1/2).

Martin-Sanchez, F. and Verspoor, K. (2014). Big data in

medicine is driving big changes. Yearbook of medical

informatics, 9(1):14.

McQuilton, P., Gonzalez-Beltran, A., Rocca-Serra, P.,

Thurston, M., Lister, A., Maguire, E., and Sansone,

S.-A. (2016). Biosharing: curated and crowd-sourced

metadata standards, databases and data policies in the

life sciences. Database, 2016.

Mons, B., Neylon, C., Velterop, J., Dumontier, M.,

da Silva Santos, L. O. B., and Wilkinson, M. D.

(2017). Cloudy, increasingly FAIR; revisiting the

FAIR data guiding principles for the European

Open Science Cloud. Information Services & Use,

37(1):49–56.

Phan, J. H., Quo, C. F., Cheng, C., and Wang, M. D. (2012).

Multiscale integration of-omic, imaging, and clini-

cal data in biomedical informatics. IEEE reviews in

biomedical engineering, 5:74–87.

Press, G. (2016). Cleaning big data: Most time-consuming,

least enjoyable data science task, survey says. Forbes,

March, 23.

Prinz, F., Schlange, T., and Asadullah, K. (2011). Believe

it or not: how much can we rely on published data on

potential drug targets? Nature reviews Drug discov-

ery, 10(9):712.

Pundir, S., Martin, M. J., and O’Donovan, C. (2017).

Uniprot protein knowledgebase. Protein Bioinformat-

ics: From Protein Modiﬁcations and Networks to Pro-

teomics, pages 41–55.

Razick, S., Mo

cnik, R., Thomas, L. F., Ryeng, E., Drabløs,

F., and Sætrom, P. (2014). The eGenVar data manage-

ment system-cataloguing and sharing sensitive data

and metadata for the life sciences. Database, 2014.

Silva, L. B., Trifan, A., and Oliveira, J. L. (2018). Montra:

An agile architecture for data publishing and discov-

ery. Computer methods and programs in biomedicine,

160:33–42.

Trifan, A. and Oliveira, J. L. (2018). A FAIR market-

place for biomedical data custodians and clinical re-

searchers. In 2018 IEEE 31st International Sympo-

sium on Computer-Based Medical Systems (CBMS),

pages 188–193. IEEE.

Vaccarino, A. L., Dharsee, M., Strother, S. C., Aldridge,

D., Arnott, S. R., Behan, B., Dafnas, C., Dong, F.,

Edgecombe, K., El-Badrawi, R., et al. (2018). Brain-

code: A secure neuroinformatics platform for man-

agement, federation, sharing and analysis of multi-

dimensional neuroscience data. Frontiers in neuroin-

formatics, 12:28.

Wen, C.-H., Ou, S.-M., Guo, X.-B., Liu, C.-F., Shen, Y.-

B., You, N., Cai, W.-H., Shen, W.-J., Wang, X.-Q.,

and Tan, H.-Z. (2017). B-CAN: a resource sharing

platform to improve the operation, visualization and

integrated analysis of TCGA breast cancer data. On-

cotarget, 8(65):108778.

Wilkinson, M. D., Dumontier, M., Aalbersberg, I. J., Apple-

ton, G., Axton, M., Baak, A., Blomberg, N., Boiten,

J.-W., da Silva Santos, L. B., Bourne, P. E., et al.

(2016). The FAIR Guiding Principles for scientiﬁc

data management and stewardship. Scientiﬁc data, 3.

Yamamoto, Y., Yamaguchi, A., and Splendiani, A. (2018).

YummyData: providing high-quality open life science

data. Database, 2018.

HEALTHINF 2019 - 12th International Conference on Health Informatics

166