Examples of lessons learned in Omics-based biomarker studies from myself and colleagues in X-omics and EATRIS, for an audience of biobankers, researchers and diagnostic/clinical chemistry experts.
2023-11-09 HealthRI Biobanking day_Amsterdam_Alain van Gool.pdf
1. Lessons learned from Omics biomarker studies
Health-RI Nationale Biobanken & Collecties dag
9 nov 2023
Amsterdam
Prof. Alain van Gool
Professor Personalized Healthcare
Translational Metabolic Laboratory
Radboudumc Technology Centers
Netherlands X-omics Initiative
Radboud Healthy Data Programme
EATRIS Biomarker Platform
2. 2 Health-RI Nationale Biobanken & Collecties dag, Amsterdam, 8 nov 2023, Alain van Gool
A short story: A biomarker validation study
• Within Schering-Plough 4 Lead Optimisation programs in ERK pathway (2009)
• Need for biomarker that indicated downstream effects of drugs:
1. Inhibition ERK pathway (pharmacodynamic, mechanistic)
2. Tumor inhibition (efficacy, clinical)
3. Blood-based (circulating)
• Extensive transcriptomics profiling: IL-8 protein as promising candidate biomarker
Data for RAFi #4
RAFi
MEKi
ERKi
RAFi
#1
RAFi
#2
RAFi
#3
RAFi
#4
MEKi
#1
MEKi
#2
ERKi
#1
3. 3 Health-RI Nationale Biobanken & Collecties dag, Amsterdam, 8 nov 2023, Alain van Gool
Validation study to confirm IL-8 as biomarker in melanoma
Literature
{Yurkovetsky, et al. Clin Cancer Res, 2007}
Study objectives:
• Confirm elevated IL-8 protein in melanoma
• Develop IL-8 immunoassays for use in clinical
of BRAFi, MEKi and ERKi
4. 4 Health-RI Nationale Biobanken & Collecties dag, Amsterdam, 8 nov 2023, Alain van Gool
Validation study to confirm IL-8 as biomarker in melanoma
Stage 1 Stage 2 Stage 3 Stage 4
H&E staining; 20x
1. Genetic analysis for BRAFV600E/D mutation in genomic DNA from tissue
2. IL-8 mRNA analysis in tissue samples by in situ hybridisation using bDNA probes
(multiplexing with 12 ERK pathway response transcripts)
3. IL-8 protein analysis in tissue samples by immunohistochemistry (in parallel with 4 other
ERK pathway response proteins, Ki67, Tunnel)
4. IL-8 protein analysis in matching plasma and serum by IL-8 immunoassay
(3 formats: ELISA, Luminex, Mesoscale; singleplex and multiplex; 2 IL-8 protein standards)
OK
OK
?
OK (also WT BRAF?)
59 melanoma samples (tumor tissue (ffpe) + matching serum & plasma, stage I-IV, from two independent
biobanks) + 40 healthy serum & plasma samples
5. 5 Health-RI Nationale Biobanken & Collecties dag, Amsterdam, 8 nov 2023, Alain van Gool
Validation study to confirm IL-8 as biomarker in melanoma
Literature
{Yurkovetsky, et al. Clin Cancer Res, 2007}
Own data
{Unpublished, 2010}
Cause?
(6 months, 5 fte, USD 1.000.000)
6. 6 Health-RI Nationale Biobanken & Collecties dag, Amsterdam, 8 nov 2023, Alain van Gool
Lessons learned?
Particularly for this case:
1. Know sample history
• IL-8 protein appeared sensitive to freeze-thawing
• History of our samples = unknown?!
2. Know all relevant information from the source (patient)
• Tumor load may have been too low for our patients
• Tumor load of our donors = unknown?!
3. We need to improve experimental reproducibility!
• Share lessons learned and do’s / don’t’s
• Share irreproducible results
• Do this together: share burden, increase power, ensure better data
• Avoid frustration and loss of time / money
{Stephan Nierkens, UMC Utrecht}
6
7. 7 Health-RI Nationale Biobanken & Collecties dag, Amsterdam, 8 nov 2023, Alain van Gool
Irreproducibility of data
{Freedman et al, PLOS Biology, 2015}
Half of published data cannot be reproduced!
Also see work from
Prof John Ioannidis (Stanford)
8. 8 Health-RI Nationale Biobanken & Collecties dag, Amsterdam, 8 nov 2023, Alain van Gool
Categories of errors leading to irreproducibility
{Freedman et al,
PLOS Biology, 2015}
From own experience:
Unawareness of critical steps in:
Pre-analytics:
• Sample generation
• Sample isolation
• Sample handling
• Sample storage
• Sample preparation
• Standards preparation
Analytics:
• Sample analysis
• Standards analysis
• Raw data processing
• Data stewardship at source
Post-analytics:
• Data integration and analysis
• Data interpretation
• Reporting
9. 9 Health-RI Nationale Biobanken & Collecties dag, Amsterdam, 8 nov 2023, Alain van Gool
But … crap data will remain crap data
even if made FAIR !
Big hopes for
10. 10 Health-RI Nationale Biobanken & Collecties dag, Amsterdam, 8 nov 2023, Alain van Gool
{van Gool et al, Nature Reviews Drug Discovery, Apr 2017}
Time for quality, not quantity !
10
11. The Netherlands X-omics Initiative
• National Roadmap Large Scale
Research Infrastructure
• 2018-2028
• €40M, partially funded by NWO
Objectives:
1. Advance X-omics technologies far beyond
state-of-the-art
2. Realize a genuinely integrated X-omics
infrastructure in NL
3. Develop and demonstrate impact of robust X-
omics analysis
Coordinator:
Alain van Gool
12. Example: population screening
▪ ERGO (Erasmus Rotterdam Gezondheid Onderzoek)
▪ 20.000 people 40+ years old in Rotterdam Ommoord
▪ Study: large scale test of 10.000 samples from ERGO biobank
▪ Glucose, cholesterol, hormones, electrolytes (Na, K)
▪ Analysis using clinical optimized protocols and auto-analysers
▪ Results: Na levels all a bit too high (?)
▪ Retest 2.000 samples: normal Na levels
▪ Probable reason after several test-retests:
▪ Electrolytes were usually measured last in the row of multiple tests (but not always), evaporation
yielded higher concentrations
▪ The order of measurement in the auto-analyzer affected the value obtained
▪ Consequence
▪ Frustration, loss of time and of 15.000 aliquots (some 20 years old and rare)
Prof Arfan Ikram
ErasmusMC
13. Example: Metabolomics of tissue of mouse models
• Question: Can we study tissue-specific pathophysiology in mice model
using metabolomics?
Example: sarcopenia and metabolic syndrome
• Observation: Variation between tissue samples from same mice ?!
• Probable reason:
The order and time used for harvesting and processing
of organs affected the metabolomic profiles obtained.
• Consequence: frustration, loss of animals and time
• Measure: Check adenylate energy charge
He Y et al, Metabolites (2022) 12 (8): 742
Prof Thomas Hankemeier
Leiden University
14. 14 Health-RI Nationale Biobanken & Collecties dag, Amsterdam, 8 nov 2023, Alain van Gool
Example: multi-cohort metabolomics study
• EU H2020 project on multi-omics biomarker identification
in primary and endocrine hypertension
• 12 sites collected 337 human plasma samples
• Metabolomics analysis @Radboudumc by NMR and LC-MS
(PhD student Nick Bliziotis)
• PCA analysis after runs: clear bias to the collection sites
• LC-MS identification of differential metabolites
• Probable reasons:
• Contaminants from collection tubes
• Precentrifugation delays
• Strong message and recommendations in
Nick’s PhD thesis
• Have we learned lessons as a field?
Collection sites
15. 15 Health-RI Nationale Biobanken & Collecties dag, Amsterdam, 8 nov 2023, Alain van Gool
Things to do still in Omics biomarker studies
• Further improve omics technologies (cell, individual, population)
• Increase awareness of variabilities and their impact in research
• Further increase analytical robustness of omics platforms
• Fully implement FAIR-data at source
• Increase capabilities to combine omics with other data
• Design and build improved (AI) models for complex data analysis
• Implement from research in health and healthcare (TRL 1 → 9)
• Share and use lessons learned - do’s and dont’s (scalable !)
www.slideshare.net/alainvangool
16. 16 Health-RI Nationale Biobanken & Collecties dag, Amsterdam, 8 nov 2023, Alain van Gool
Some do’s and dont’s
In lab protocols
• Test variability in each step
• 5 min = 5:00 or >3 min?
• 37°C = 37,00 or between 36 and 38°C?
• Clearly mark the critical step!
• Describe analytical standard well
• Specific source, lot, date
• Include test of several standards !
• Test sensitivity of sample for your test
• Freeze-thawing
• Sensitivity at 20°C, 4 °C and storage
• Effect of contaminants
In teaching
• Allow people to make mistakes (students!)
• Communicate failures as well + reasons why
• Overall message: First search and think, then do
In data processing
• Software / database: describe specific source, version,
date
• Hardware: describe type, version + adaptations (!)
• Follow FAIR data stewardship guidelines
In collaborations
• Share protocols, standards and samples
• Redo each other’s analysis
• Improve where possible
Proteomics labs Radboudumc and ErasmusMC
Charissa Wjnands, Clin Chem Lab Med 2023
17. 17 Health-RI Nationale Biobanken & Collecties dag, Amsterdam, 8 nov 2023, Alain van Gool
Sharing lessons learned
70 manuscripts from experts in pharma,
diagnostics, clinic, technology
1. What is a biomarker and their role in drug
development?
2. Biomarkers in preclinical sciences
3. Biomarkers in translational sciences
4. Biomarker-informed clinical trials
5. The road ahead in precision medicine
6. Lessons from the past and pioneers of the future
7. Emerging technologies
8. The next frontiers in therapeutic target areas
9. Lessons learned and what’s next?
History – Current status – Future
Lessons learned
19. Data Stewardship
& Integration (WP2)
Multi-omics
technologies (WP1)
Genomics
Epigenetics
Transcriptomics
Proteomics
Metabolomics
Quality Assessment
(WP3)
Data
Generation
Data
Curation
Quality
Assessment
Data
Variability
Multi-omics Toolbox
Academia & Industry
Pilot Access
• SOPs
• Guidelines for best practices
• Reference materials
• Quality parameters
• Data analytical tools
• Criteria for reference values
• Troubleshooting guidelines
• Repository of multi-omics data
An open access WEB
resource, containing:
Enable high-quality research in PM, including
accelerating the implementation of accurate
patient stratification and treatment management
EATRIS-Plus Key Scientific Objective
flagship project
20.
21. Supporting translational researchers
▪ Translational challenge:
Poor levels of technological
and analytical harmonisation
▪ What does MOTBX offer to
help overcome this
challenge?
• Standardised protocols
for omics measurements
and data processing
• Guidelines for quality
control and assessment
22. Supporting translational researchers
▪ Translational challenge:
Poor data stewardship and
compliance to the FAIR
(Findable, Accessible,
Interoperable, Reusable)
principles
▪ What does MOTBX offer to
help overcome this
challenge?
• Guidelines on FAIR data
stewardship, resources
about data and metadata
standards, and data
management tools
23. 23 Health-RI Nationale Biobanken & Collecties dag, Amsterdam, 8 nov 2023, Alain van Gool
Different behaviour life scientists is needed to improve quality
Adopt a lesson learned
somewhere else?
24. 24 Health-RI Nationale Biobanken & Collecties dag, Amsterdam, 8 nov 2023, Alain van Gool
Different behaviour life scientists is needed to improve quality
“Every-one wants to innovate,
but no-one wants to change”
Bas Bloem (Radboudumc)
on clinical translation
“' We are really good at innovation
but bad in scaling”
Prins Constantijn van Oranje
at Health-RI conference 2021
“We should move from Proudly invented here
to Proudly copied from“
Alain van Gool
at DTL Partner Event 2022
25. Acknowledgements
Hans Wessels
Jolein Gloerich
Dirk Lefeber
Karlien Coene
Purva Kulkarni
Gadi Armony
Richard Rodenburg
Bert van den Heuvel
and others
Marcel Nelen
Albert Heck
Thomas Hankemeier
Peter Bram ‘t Hoen
Daniella Kasteel
Jessie Smits
and others
Collaborators/funders
Helger Ijntema
Alexander Hoischen
Lisenka Vissers
Christian Gillisen
Janneke Weiss
Han Brunner
alain.vangool@radboudumc.nl
www.radboudumc.nl/en/people/alain-van-gool
www.slideshare.net/alainvangool
Dapha Habets
Irene Keularts
Marek Noga
and others
Gary Kruppa
Pierre-Olivier Schmit
Dennis Trede
and others
Toni Andreu
Florence Bietrix
Emanuela Oldoni
Laura Garcia Bermejo
Andreas Scherer
Eliis Keidong
and others
Hans Jacobs
Peter-Bram ‘t Hoen
Anna Niehues
Xiaofeng Liao
Casper de Visser
Junda Huang
Translational Metabolic Laboratory
Human Genetics Nijmegen
Center for Molecular and Biomolecular Informatics