The document summarizes the NCI Childhood Cancer Data Initiative (CCDI). The CCDI aims to build a community focused on pediatric and adolescent cancer by improving treatment, quality of life, and survivorship through learning from every child. It discusses establishing the CCDI structure and working groups, and outlines three pillars of the CCDI: the Molecular Characterization Initiative, the National Childhood Cancer Cohort, and the National Childhood Cancer Registry. The Molecular Characterization Initiative launched in 2022 to expand access to molecular sequencing for childhood cancers.

1. Learning from Every Child :
The NCI Childhood Cancer
Data Initiative
December 8, 2023
The Wake Forest Center for Artificial
Intelligence Research
@wakibbe
Warren A. Kibbe, Ph.D.
Vice Chair & Professor of
Biostatistics and Bioinformatics
Duke University School of Medicine
warren.kibbe@duke.edu

2. 2
NCI Childhood Cancer Data Initiative (CCDI) –
Building A Community
Childhood and AYA cancers are rare diseases: only 4-5% of diagnosed cancers
In the US: About 16,000 cases for children from birth to 19 years of age
About 80,000 cases for young adults from 20 to 39 years of age
Acute need for datato support research
CCDI aims to build a
community focused on Pediatric and AYA Cancer

3. CCDI
improves
treatment
quality
of life
s u r v i v o r s h i
p
by
learnin
g from
every
child

4. Personal & Professional Background
•PhD in Chemistry at Caltech, Postdoc in molecular genetics of
RAS at the Max-Planck-Institute for Biophysical Chemistry
•Cancer research for 25+ years - cancer informatics, data
science, healthcare – open science, open data advocate
•Feinberg School of Medicine at Northwestern for 15+ years
•Director NCI CBIIT 2013-2017; NCI CIO 2013-2017; Acting NCI
Deputy Director for Data Science 2016-2017
•At Duke since 2017
•Currently on IPA with NCI to co-lead the CCDI
•Lost three grandparents to cancer, father to cancer in 2019

5. Disclaimer
The views I express are mine and not endorsed by Duke or NCI.

6. The importance of Open Science
Calls for greater transparency and ‘open data access’ in clinical research continue
actively.
“Open science is the movement to make scientific research, data and
dissemination accessible to all levels of an inquiring society”*
Open Science Project**: “If we want open science to flourish, we should raise our
expectations to: Work. Finish. Publish. Release.”
FAIR Principles: Findability, Accessibility, Interoperability, and Reusability***
TRUST Principles: Transparency, Responsibility, User focus, Sustainability and
Technology****
* https://www.fosteropenscience.eu/resources
** http://openscience.org/
*** https://www.nature.com/articles/sdata201618
****https://www.nature.com/articles/s41597-020-0486-7

7. 7
cancer.gov/CCDI #data4childhoodcancer
Build a strong base: Progress requires
data from many sources that is
connected and easy to access.
Assemble better data: Complete
data sets are needed to understand
each type of cancer.
Make data easy to use: More
thoughtful tools for analyzing data will
help answer important questions.
Improve treatments and outcomes:
Data is the foundation that informs
new treatments and improves lives
faster.

8. “Focus on the critical need to collect, analyze, and share data better in order to maximize NCI’s ongoing
investment in pediatric and AYA cancers and survivorship. Tissue samples from patients with cancer in this age
group are critically limited and a valuable resource. The overall goal of the CCDI is not simply to generate more
data, but to build processes that transform data into knowledge that moves the field forward in meaningful
ways. The CCDI supports the wider pediatric cancer community’s goal of maximizing the stated goal of
“learning from every patient” so that ultimately those patients, survivors and their families can materially
benefit in terms of higher cure rates and improved long-term health outcomes.”
Aggregate and
generate broad
categories of data
Develop
infrastructure
Engage with experts
Empower patients
& families
Ensure appropriate
policy and funding
Develop strategy for
survivorship
Ensure diverse
patient
representation
Enable improved
patient outcomes
and treatment
Categories of Recommendations
BSA CCDI Working Group Report
24 recommendations
WG Report
https://deainfo.nci.nih.gov/advisory/bsa/sub-cmte/CCDI/CCDI%20BSA%20WG%20Report_Final%20061620.pdf

9. Ad Hoc Working Group in Support of the CCDI
Kevin Shannon
(Co-Chair)
UCSF
Otis Brawley
(Co-Chair)
Johns Hopkins
Peter Adamson
Sanofi
Tom Curran
Children’s
Mercy
James Downing
St. Jude
Julie Guillot
Leukemia &
Lymphoma
Society
Amanda
Haddock
Dragon Master
Foundation
Samuel
Volchenboum
U Chicago
John Maris
CHOP
Andrew Kung
MSKCC
Warren Kibbe
Duke
Andrea Hayes-
Jordan
UNC
Katherine Janeway
DFCI
Douglas Hawkins
Seattle Children’s

10. • Adapt treatments,
diagnostics, and
prevention
strategies as
knowledge is
gained
• Identify exceptional
responders
• Understand
molecular
landscape of rare
cancers
• Use germline for
predisposition
studies
• Aggregate, federate
• Data resource
inventory
• Identifiers for
biospecimens
• Develop NCCR
• Tools for translation
• Harmonize clinical
and research
terminologies
• AL and ML
approaches
• Simplify data access
• Bar coding for data
• Aggregate 6 broad
categories of data
• Existing molecular
• New “ideal” data
types
• Preclinical
• Off-trial patient
data
• Develop national
strategy for
molecular
characterization
• Cancer model data
Aggregate and
generate data
Develop
infrastructure
Enable improved
patient outcomes
and treatment
• Allow patients to
access their data
• Creative ways to “opt
in”
• Engage with all
stakeholders
Empower patients &
families
BSA CCDI Working Group Recommendations

11. • Make data
representative of full
spectrum of patients
• Consent diverse
populations
• Develop long-term
strategy for tracking
patients across
lifetime
• Set expectations for
data sharing
• Commit to data
sharing thru policy
and funding
• Convene experts for
molecular
characterization
• Convene experts for
NCCR
• Engage with all
stakeholders
Engage with experts
Ensure appropriate
policy and funding
Develop strategy for
survivorship
Ensure diverse
patient
representation
BSA CCDI Working Group Recommendations

12. Establish
Goals of
Initiative
Identify
Gaps,
Needs,
Research
Priorities
Develop
Structure to
Meet Goals,
Addressing
Gaps
Set
Milestones &
Assessment
Metrics
Begin FY22-FY24
Concept
Development
Process
State of
the Union
Kick-Off
Symposium
CCDI Structure
(BSA/NCAB)
CCDI Working Groups
& CCDI Leadership
BSA Working
Group Report
2019 - 2021
NCI: Leverage existing programs, funding mechanisms,
datasets for short-term solutions to implement vision
2022 - 2029
Issue Funding
Opportunities
Mix of funding based on NCI
strategic planning
CCDI Program Establishment

13. 13
cancer.gov/CCDI #data4childhoodcancer
The Three
Pillars of
CCDI

14. 14
Pediatric/AYA data from multiple sources
Improved understanding of why some cancers
develop resistance or don’t respond to treatment
Generation of new ideas for
intervention
Culture change towards improved
collaboration and data sharing
Development of new research
and analytical tools
CCDI
NCI - COG
Pediatric MATCH

15. Patient Data Touchpoints
Diagnosis
Confirmation / Refinement
Clinical presentation, biopsy, imaging
Molecular Sequencing,
proteomics, etc.
Decision-making / Clinical Care
Molecular Tumor Board, Review notes,
treatment decisions
Longitudinal Follow-up
Outcomes
Long-term outcomes
Short-term outcomes
Treatment
Treatments, procedures, adverse events
Secondary
Cancer
Recurrence
Progression
Data Touchpoint
Epidemiology / Population Sciences Data –
Familial data, environmental, registry, population studies, disease cohorts
Screening, Diagnosis, Treatment

16. Our ability to generate biomedical
data continues to grow in terms of
variety and volume
Current sources of data
molecular genome pathology imaging labs notes sensors
icons by the Noun Project

17. Diagnosis
Molecular Characterization
Molecular Tumor Board
Treatment
Follow-on Cancer
Recurrence
Long-term follow-up
Outcomes
Federated
Data
Ecosystem
Cancer Data “Scorecard”

18. Diagnosis
Molecular Characterization
Clinical Decision,
Molecular Tumor Board
Treatment
Follow-on Cancer
Recurrence
Long-term follow-up
Outcomes
Federated
Data
Ecosystem
Cancer Data “Scorecard”
EHRs, LIMS, Radiology, Pathology Systems
Tumor Board Systems, Hospital
Systems, EHRs, Clinical Trial Matching
EHRs, CDMS, Adverse Events, CTMS
EHRs, LIMS, Radiology, Pathology Systems
EHRs, CDMS, CTMS
EHRs, Registries
EHRs, LIMS, Radiology, Pathology Systems
LIMS, Sequencing Systems

19. 19
CCDI Initiatives

20. 20
CCDI Molecular Characterization Initiative
https://www.nih.gov/news-events/news-releases/nih-launches-program-offer-molecular-characterization-childhood-cancers
Launched March 21, 2022
First patient enrolled
April 2022

21. 21
Childhood Molecular Characterization Initiative
• Expand access to comprehensive molecular sequencing as a step towards the
goal of reaching all children with pediatric cancer
• Develop NCI-recommended guidelines for clinical and molecular data
collection as part of standard of care
• Create a comprehensive, harmonized, and integrated database of clinical,
genomic, and phenotypic data for research
• First patient was enrolled in April 2022!
Clinical Service: Diagnostic clinical molecular
characterization services for patients who might
not otherwise have access to them
Data to be collected (CLIA certified)
• DNA: CLIA WES or NGS targeted panel
• Fusion panel
• Methylation: CLIA DNA Methylation array
• Clinical annotation
Research Discovery: Molecular characterization to learn
more about disease subtypes and rare cancers
Data to be collected (in additional to clinical/seq data on
selected populations)
• WGS/deep molecular (DNA) profiling
• RNAseq
• Longitudinal data

22. 22
MCI
 As of November 2023 – cumulative enrollment surpassed 2,000
 1,965 patients from ~180 institutions have a sample shipped to the
Biopathology Center
o CNS: 1,397; soft tissue sarcoma: 429; rare tumors: 139
 Phs002790: Monthly molecular characterization data releases (1,697
participants).
o Deidentified clinical reports also available in JSON format (converted from PDF).
 COG releases clinical data every 6 months
 Clinical reports currently support 5 clinical trials:
 Timely return of results has enabled trial enrollment in short windows.

23. Clinical
Sequencing at IGM
(CLIA)
 Targeted exome seq
 Archer fusion
 EPIC methylation
array
 Clinical/demographic
PI: Elaine Mardis
PO: Malcolm Smith
Research-Grade
Characterization
(non-CLIA)
 WGS
 RNA-seq
 Proteomics
 Metabolomics
 Other
 Clinical
COG Project:
EveryChild
(CNS tumors,
sarcomas, or rare
cancers)
PI: Doug Hawkins/ Dx
Chairs
PO: Malcolm Smith
MyPART
Refractory or relapsed
cancer types (adult &
pediatric) with rare
cancers (DNA/RNA)
PIs: Brigitte
Widemann, Karlyne
Reilly
CCDI Molecular Characterization Initiative: Clinical Pipeline
CCDI Data
Ecosystem (AWS)
Deposit clinical
and genomic
data
Return of CLIA
genomic results
to patients and
providers
CCDI Molecular
Characterization
Pipeline
Protocol Enrollment
(Goal: determine best strategy
to treat & learn from each child)
CCDI Funded
PI: Jaime Guidry Auvil, Tony Kerlavage
PO: Subhashini Jagu
Specific to
Enrollment
Other NCI Studies
Patient enrollment, CLIA
tissue processing,
clinical data collection,
mechanism to return
patient results
Additional
Sources of Tissue
Samples & Clinical
Data (NEW)
The pipeline began accepting samples from PEC CNS tumors in April 2022

24. Clinical
Sequencing at IGM
(CLIA)
 Targeted exome seq
 Archer fusion
 EPIC methylation
array
 Clinical/demographic
Research-Grade
Characterization
(non-CLIA)
 WGS
 RNA-seq
 Proteomics
 Metabolomics
 Other
 Clinical
Kids First X01s
CBTN (CNS tumors;
PI: Adam Resnick),
Sarcomas (PI:
Adam Shlein)
PO: Malcolm Smith
PDX Models
PI: PiVOT
Consortium
PO: Malcolm Smith
Pediatric MATCH
Dx/Normal tumors
PI: Will Parsons/Katie
Janeway
PO: Nita Seibel
Childhood
Cancer Survivor
Study (CCSS)
(Secondary Cancers)
PI: Greg Armstrong
PO: Nita Seibel
CCDI Molecular Characterization: Research Pipeline
CCDI Data
Ecosystem (AWS)
Deposit clinical
and genomic
data
Return of CLIA
genomic results
to patients and
providers
CCDI Molecular
Characterization
Pipeline
Protocol Enrollment/
Sample Collection
(Tissue & Clinical
Data)
**CCDI Funded
PI: Broad Institute,
Hudson Alpha/St. Jude
PO: Malcolm Smith
CCDI Funded
PI: Jaime Guidry Auvil, Tony Kerlavage
PO: Subhashini Jagu
The pipeline released the first bolus of data for research use in November 2022

25. Clinical Sequencing
(CLIA)
 Targeted exome seq
 Archer fusion
 EPIC methylation array
 Clinical/demographic
Research-Grade
Characterization
(non-CLIA)
 WGS
 RNA-seq
 Proteomics
 Metabolomics
 Other
 Clinical
CCDI Protocol
Pediatric Cancer Molecular & Clinical Characterization: CCDI Protocol
CCDI Data
Ecosystem
Deposit clinical
and omic data
Return of CLIA
genomic results
to patients and
providers
Patient remains
on CCDI
Protocol; may be
enrolled in sub-
studies or
clinical trials
COG
Project:
EveryChild
MyPART
Other NCI
Studies
Additional
Sources
Kids First
X01s
PDX Models
Pediatric
MATCH
CCSS
Longitudinal follow-up
for select patient
cohorts
CCDI Clinical
Characterization
Pipeline
Comprehensive clinical characterization
 Etiology, family history, medical history
 Clinical, PRO, imaging
 Pathology
 NIH Rare Tumor clinics
 National tumor Board
 Genetic counseling
Core set clinical characterization
Future State/Vision:
• A single study acts as an entry point clinical and research-
grade sequencing
• The study collects a comprehensive baseline and longitudinal
dataset for all patients and an enriched dataset

26. 26
• Gather data from every child diagnosed with cancer in the United States
• It will:
 Capture the cancer care trajectory of children and AYAs, including care
provided outside of COG and other networks, to identify gaps and
disparities in care and outcome
 Track biospecimen availability
 Provide access to data from underserved patients
 Provide for consistent research consent
 Allow for long-term follow up of childhood cancer patients
• A critical component of this effort will be the National Childhood Cancer
Registry (NCCR)
CCDI National Childhood Cancer Cohort

27. The National Childhood Cancer Registry (NCCR)
27
• Slides from Dr. Lynne Penberthy, NCI

28. The National Childhood Cancer Registry (NCCR)
28
• NCCR is a centralized infrastructure that brings together existing data on
all cancer patients ages 0 to 39
• Core data are derived from cancer registries
• Data from registries are extended and expanded to include additional relevant
information such as
• Detailed treatment
• Genomic characterization of the tumor
• Risk of recurrence
• Risk of subsequent primary cancers
• With a goal to capture the complete trajectory of care from diagnosis
throughout the patient’s life
• Currently includes patients representing 77% of all childhood
cancers with 21 state cancer registries participating

29. What is a
cancer
registry?
• Cancer registry data provide the base for the
NCCR
• Registries:
• collect, store, and manage data on every person with
cancer from diagnosis until death
• are population based (capture all cancers within a
defined geographic area)
• have the legal authority to capture data on every
cancer
• require health care providers to report information on
cancer patients to state registries
29

30. The National
Childhood
Cancer
Registry
Components
30
Routine linkages will be performed centrally with
external data sources including:
• Complete registry abstracts for each cancer
case (1995-2019+)
• Survival data from National Death Index (NDI)
& State vital records
• Residential history data
• essential to enable data to be linked over
time for treatment, recurrence,
subsequent cancers or adverse effects
• Virtual Pooled Registry (VPR)
• A national system of virtual connections
for all state registries the supports
• de-duplication (many pediatric
patients receive care across state
boundaries) and
• allows capture of subsequent cancers
as patients age

31. New NCCR data sources
Data sources currently used
• Birth
Records,
Blood spots)
Selected
state vital
records
•Long-term
follow-up
center, AACCR
Children’s
Oncology
Group
SEER Cancer
Registries
Leveraging Existing
SEER*DMS servers
holding PII for
Secure Data
Platform
Selected State
Cancer Registries
(including TX, TN, PA, IL, NJ,
OH, FL)
Instance of DMS*Lite
For each registry to hold
PII for linkages
NCCR
Database***
Combines
de-identified
data submitted
from
participating
registries plus
linked data
from additional
sources
Conceptual Framework: National Childhood Cancer Registry
***NCCR Database– holds de-identified childhood cancer
patient data submitted from participating registries.
Infrastructure to support research on childhood cancers 31

32. Why are the data from NCCR important: Providing a Report Card -
Leukemia as an example*
*https://seer.cancer.gov/statistics/nccr/
• Registries provide de-identified
data on every patient with cancer
in a defined geographic region
• These data allow us to
• Monitor progress (report
card) of clinical care on
outcomes
• Help us identify groups of
patients who may not be
experiencing the benefit of
the new treatments by age,
racial and ethnic subgroups,
or geographic area
• Similar trends are being developed
for survival and mortality
32

33. 33
NCCR Explorer
https://nccrexplorer
.ccdi.cancer.gov

34. 34
CCDI Data Ecosystem

35. 35
• Designed to federate data from multiple children’s cancer institutions
and community-based and NCI-supported childhood/AYA data
resources, featuring:
 Patient-level data from all available sources
 Easy access to data to enable deep analytics
 Supports interoperability among existing data resources and with tools
and other resources for use by researchers
 Provide a central portal to find and analyze childhood/AYA cancer data
CCDI Childhood Cancer Data Platform

36. 36
Establish a Federated Pediatric Cancer Data Ecosystem:
 Underlying data science
infrastructure
 Enhanced cloud-computing
 Services linking clinical, image, &
molecular data
 Standards & tools for data
interoperability
 Data repositories (e.g. Pediatric
Preclinical Data Commons)
 Linked data (National Childhood
Cancer Registry) Genomics Data
Proteomics
Data
Cancer Models
Molecular
Imaging Data
Clinical
Trials
Treatments
Patient Outcomes
Cohort Studies
Electronic
Health Records
Preclinical
Data
Immuno-oncology
Data
Demographics
Clinical
Characterization
Discovery Science  Clinical Studies/Care 

37. Childhood Cancer Data Catalog
CCDI Annual Symposium
A guide to finding childhood cancer programs, registries, and repositories
Slides thanks to Dr. Patrick Dunn, Frederick National Laboratory

38. 38
Pediatric Cancer Data Resources

39. 39
ACatalog of Data Resources for Childhood Cancer Research
A guide to
relevant
research
programs,
clinical trials,
data sets and
repositories to
support the
RACE for
Children Act
*
https://datacatalog.ccdi.cancer.gov

40. 40
cancer.gov/CCDI #data4childhoodcancer
CCDI Data Ecosystem Components: Connecting the Data
Primary databases
 Cancer Research Data Commons (CRDC)
 Childhood Cancer Clinical Data Commons
(C3DC)
 National Childhood Cancer Registry
(NCCR)
Knowledge Bases
 Childhood Cancer Data Catalog
 Molecular Targets Platform
 Data Inventory
 Participant Index
Curation & Harmonization
 Data Coordination Center – assist with data
submission and harmonization

41. 41
cancer.gov/CCDI #data4childhoodcancer
Access to CCDI Data & Tools
ccdi.cancer.gov
 The CCDI Hub is an entry point for researchers,
data scientists, and citizen scientists looking to
use and connect with CCDI-related data.
 It provides information and direct links to CCDI
platforms, tools, and resources, along with
additional technical information.
 CCDI Hub 2.0 release
 Integrate with CCDI Data Inventory; leverage
existing Bento design, customized for CCDI.
 e.g., https://dataservice.datacommons.cancer.gov/#/data
 Sharing CCDI clinical metadata and linkages
to sample data.

42. 42
cancer.gov/CCDI #data4childhoodcancer
PedcBioPortal at KidsFirst

43. 43
cancer.gov/CCDI #data4childhoodcancer
 phs002790.v1.p1 (CNS, STS, rare tumors - MCI)
 Case count: 1,552; Sample count: 4,377
 phs002599.v1.p1 (Acute myeloid leukemia - OHSU)
 Case count: 129; Sample count: 1347
 phs002504.v1.p1 (Juvenile myelomonocytic
leukemia - UCSF)
 Case count: 188; Sample count: 388
 phs002620.v1.p1 (Solid tumors - MSK)
 Case count: 114; Sample count: 326
 phs003111.v1.p1 (High-risk neuroblastoma - MSK)
 Case count: 130, sample count: 488
 phs002518.v1.p1 (All cancer types - USC)
 phs002518.v1.p1 (All cancer types - USC)
 Case count: 1036 Sample count: 2240
 phs002529.v1.p1 (Solid tumors & leukemias -
CMRI/KUCC)
 Case count: 194; sample count: 373
 phs002517.v1.p1 (Brain & other solid, hematologic
malignancies - CHOP)
 Case count: 1031; sample count: 3,225
 phs000720.v4.p1 (Rhabdomyosarcomas whole-
slide hematoxylin and eosin-stained images from
COG trials ARST0331, ARST0431, D9602, D9803,
and D9902)
 Will be available through NCI’s Imaging Data
Commons and listed in the Childhood Cancer Data
Catalog
CCDI Data Sets

44. 44
CCDI Data Access
 For the CCDI studies, genomic data is hosted in
the Cancer Data Service (CDS), which is a data
repository under the Cancer Research Data
Commons infrastructure.
 dbGaP maintains a list of the subject IDs, sample
IDs, and consents.
 Accessing controlled-access data and
clinical/phenotypic files requires authorization
through dbGaP.
 Users can analyze CCDI data on the Cancer
Genomics Cloud through the Cancer Data Service
Explorer.
 Here is a tutorial on how to import CDS
data: docs.cancergenomicscloud.org/docs/import-
cds-data
CCDI_CGC_Data_Access_Inst
ructions:
datacatalog.ccdi.cancer.gov/CC
DI_CGC_Data_Access_Instruct
ions_1.0.pdf

45. 45
Data Federation Demonstration Project
Goal: Facilitate large-scale analytic research
through heterogeneous data aggregation.
 Make deidentified participant-level data (non-
PHI/PII) findable across the sources, enabling
the creation of a virtual cohort (without moving
or warehousing data).
Status Update:
 Developed scientific use cases to drive
requirements for data and an API.
 Defining a minimal set of demographic and
clinical phenotype data to allow queries across
the CCDI Federation data resources.
 Drafting API requirements for the CCDI
Federation.

46. 46
Childhood Cancer Clinical Data Commons (C3DC)
Goal: Allows researchers to search for
participant-level data collected from multiple
studies to create synthetic cohorts.
Status Update
 Working on establishing the reference set of
Common Data Elements (CDEs) that will
constitute a pediatric cancer clinical data
standard.
 Created C3DC data model in GitHub.
 Tentative MVP release Q4 of 2023.
How do I find the CDEs?
cadsr.cancer.gov/onedata/Home.jsp

47. 47
cancer.gov/CCDI #data4childhoodcancer
Childhood Cancer Data Initiative Participant Index (CPI)
Goal: Link participants’ data from the Federation and the
CCDI projects.
Status Update
 COG has provided an endorsement that authorizes
institutes to share the COG IDs with NCI.
 Received COG ID - USI pairs for MCI participants.
 Will facilitate crosswalk of CBTN data to overlapping MCI
participants
 Received Kids First Study IDs mapped to USIs.
 Working with other institutions to acquire mapped IDs
 Identifying institutions to participate in PPRL process.
 Working with NCCR
 Defining the technical architecture for the database.

48. 48
learn
from
every
child.

49. 49
engage with CCDI
• Visit the CCDI website to learn more
(on the NCI website)
• Review the BSA CCDI Working Group
final report
• Receive email updates from NCI on the
CCDI
• Contact the CCDI with questions about
engagement, ongoing activities, or
funding opportunities
• Look for information about webinars
starting in 2022

50. Thank you!
Thank you!

51. 51
Questions?