A key objective of the healthcare industry is to accelerate translational science, ie. the translation of scientific discovery into products and services for the benefit of our patients. The technological progress and the digital transformation offer many approaches to enhance data driven decision making. Data quality is a must have for AI outcome, and curation of data to improve machine readability is a core activity in order to enable data science.
The data quality problem of health research has been acknowledged by the European Commission and they launched the European Open Science Cloud initiative. Within this context, the FAIR Guiding Principles were published. FAIR stands for Findable, Accessible, Interoperable, and Reusable. FAIR data has become a global movement and also reached the Pharma industry.
At Bayer, we approach the FAIR data topic in three different ways: Bottom up with use case driven projects; top-down strategic initiatives to develop infrastructure, capabilities and mindset change; collaboration, i.e. cross-divisional within Bayer and across pharma in public-private consortia.
2. Making data FAIR at Bayer
/// Making data FAIR at Bayer /// A. Grebe de Barron /// Connected data London, Oct 2019
Why we
need it
What is
FAIR
How we
do it
3. /// Making data FAIR at Bayer /// A. Grebe de Barron /// Connected data London, Oct 2019
4. A transformational change in R&D productivity is required to
reverse declining trends in R&D returns across the biopharma
industry
/// Making data FAIR at Bayer /// A. Grebe de Barron /// Connected data London, Oct 2019
Deloitte: Unlocking R&D productivity / Measuring the return from
pharmaceutical innovation 2018
5. Drug development is a highly-regulated, long and costly endeavor
with a low probability of success
/// Making data FAIR at Bayer /// A. Grebe de Barron /// Connected data London, Oct 2019
Preclinical testsDrug research Clinical trials Approval
1 2 3 4
After launch
5
1 JAMA Intern. Med. 2015, 175, p. 635-638
Permanent…benefit-risk
assessment of products
15 years…from idea to
market approval
>2 billion €… investment
per asset until approval
<1%...chance for a project to
get approved
12.4 years…average time
effect. market exclusivity1
Numbers…
1 JAMA Intern. Med. 2015, 175, p. 635-638
6. Digital health revolution requires a new mental model
and offers new digital business possibilities
/// Making data FAIR at Bayer /// A. Grebe de Barron /// Connected data London, Oct 2019
The past
Innovative, effective and safe drugs
Market authorization
Focus on internal trial data, IP protection
R&D process optimization
The future
Patient centricity
Data democracy
Real world data
Data products / AI
personalized medicine, prevention
ownership, charge of care
open, collaborative partnerships
automation, science apprentice,
medical assistant
EHR
7. Mission
/// Making data FAIR at Bayer /// A. Grebe de Barron /// Connected data London, Oct 2019
grow the data workforce
be data driven not data rich
advance the digital agenda
8. Information asset 360
/// Making data FAIR at Bayer /// A. Grebe de Barron /// Connected data London, Oct 2019
a data product that has a value greater
than the sum of its parts
whereby it consists of multiple datasets
integrated for a data domain context
and unlocks insights for multiple functions
for the enterprise / division
9. for both humans and machines
FAIR is the foundation for unlocking the value of Bayer‘s data
/// Making data FAIR at Bayer /// A. Grebe de Barron /// Connected data London, Oct 2019
F A I R
Findable means to
uniquely and
persistently identify
existing data assets in
a searchable and
accessible resource
Accessible means that
data assets can be
easily retrieved upon
appropriate
authorization
does not mean open
without constraint
Interoperable allows
data to be actionable by
presenting data assets
in a formal, broadly
applicable way and
linking it to other data
Reusable clarifies the
context, meaning,
trustworthiness and
origin of data,
and how it can be used
with a clear and
accessible data usage
license
10. European Open Science Cloud (EOSC) - European Commission
What is FAIR and where does it come from?
/// Making data FAIR at Bayer /// A. Grebe de Barron /// Connected data London, Oct 2019
… aims to accelerate and support the
current transition to more effective Open
Science and Open Innovation in the Digital
Single Market.
It should enable trusted access to services,
systems and the re-use of shared scientific
data across disciplinary, social and
geographical borders.
https://www.nature.com/articles/sdata201618
11. Collaboration, especially public-private, is the key for successful
research output and innovation
/// Making data FAIR at Bayer /// A. Grebe de Barron /// Connected data London, Oct 2019
https://fairplus-project.eu/
12. Implementing a FAIR ecosystem @Bayer
/// Making data FAIR at Bayer /// A. Grebe de Barron /// Connected data London, Oct 2019 picture from www.slideshare.net
implicite digitalization
Challenges
process vs data centric value focus
fragmented investments into data
technologies and initiatives
complex static policies and regulations
incomplete data life cycle coverage
AI
explicite
Costs
set up of FAIR ecosystem/expertise
make legacy data FAIR
make data generation and integration FAIR
Create awareness, educate, change mindset,
incentivise
13. Target reference architecture to accelerate data & analytics
Implementing a FAIR ecosystem @Bayer
/// Making data FAIR at Bayer /// A. Grebe de Barron /// Connected data London, Oct 2019 picture from www.slideshare.net13
Data @ Scale Visualization, Apps &
Systems
Internal Data Sources
External Data Sources
Data Ingestion Pipelines Persistence Layer
Science@Scale
Exploration Zone
Data Robot
Commercial
R&D
Product Supply
Commercial
Product Supply
R&D
MAPV
MAPV
Data Warehouse
Data Lake
Near Realtime
Streaming
Automated Batch
Processing
Data Market
Place
Metadata & Ontologies
DataAccessLayer
Others
PID
Top
Braid
API
SQL
Stream
FAIR
ecosystem
components
SPARQL
14. Information asset 360
/// Making data FAIR at Bayer /// A. Grebe de Barron /// Connected data London, Oct 2019
a data product that has a value greater
than the sum of its parts
whereby it consists of multiple datasets
integrated for a data domain context
and unlocks insights for multiple functions
for the enterprise / division
15. Selection and definition of 360 Information Assets
is based on aligned design principles
/// Making data FAIR at Bayer /// A. Grebe de Barron /// Connected data London, Oct 2019
Must have broad relevance across the business
Reflects an interconnected portfolio of datasets
that express the future lens of our business
Connects isolated signals across the value chain
and enables multiple functions (R&D, PS, M&S)
Represents core domain datasets and is
"invariant" to normal business process changes
Reduces risk, controlling data usage & handling by
embedding security / entitlements by design
16. • Prescriber
• Sales contact
• Key opinion
leader
• Research
collaboration
• Reporter on
adverse events
• Real world
data provider
• Principle
investigator
• Trial site
Development Pharmacovigilance
MarketingResearch
Example „Healthcare professional“
360 Information Assets connect isolated signals across the value
chain and enable multiple functions
/// Making data FAIR at Bayer /// A. Grebe de Barron /// Connected data London, Oct 2019
EHR
EHR
17. Concept for a concerted implementation of information assets
/// Making data FAIR at Bayer /// A. Grebe de Barron /// Connected data London, Oct 2019
PATIENT PRODUCT CUSTOMER information
asset XYZ
FAIR culture
Sharing, governance, security, education, incentives
FAIRification process
Knowledge engineering, curation, APIs
FAIR ecosystem
IT architecture, products, services for data integration and consumption
18. Deliverables of the FAIRification process
/// Making data FAIR at Bayer /// A. Grebe de Barron /// Connected data London, Oct 2019
FAIR digital objects
data/metadata
software/code/algorithms
protocols
models
licenses
other research outputs
FAIR components
skills and investment
policies
data mgmt plans (DMPs)
persistent identifiers
standards
FAIR services
curation and stewardship
data lifecycle management
long-term preservation
file format transformation
data protection / security
handover plans for discontinued
services
19. FAIR = Findable, Accessible, Interoperable, Re-usable
Building the basic infrastructure for making data FAIR
/// Making data FAIR at Bayer /// A. Grebe de Barron /// Connected data London, Oct 2019
ontologies curation tools services
Curate
Ontologize
Standardize
Govern
FAIR datathons
bring your own data and
make them FAIR
PID
Data
Market Place
Persistent ID
(URI)
Resource
registry
Register
Share
F
A
I R
20. Acknowledgements
FAIR ladies, ontologists
Drashtti Vasant
Melanie Hackl
Johanna Völker
Olga Streibel
Marius Michaelis
Data as an asset, 360s
Manuela Schwenninger
Gökhan Coskun
Rolf Grigat
Henning Dicke
Anja Tilinski
Maria Horsch
Andy Montgomery
Tim Williamson
/// Making data FAIR at Bayer /// A. Grebe de Barron /// Connected data London, Oct 2019
MAPV information asset program
Michael Heese
Karsten Hanff
EdiSON program/FAIR data stream
Elke Hess
Angeli Möller
Saskia Schmidt-Riddle
Barbara Weidgang
Peter Borowski
DINOS
Nicole Philippi
Sebastian Lühr
Qiong Lin
Jens Scheidtmann