Clinical Working Group
Global Alliance for Genomics and Health
March 4
th
, 2014
Clinical Working Group Members
Name
Institution
Kathryn North (Co-Chair)
Murdoch Childrens Research Institute, Melbourne, Australia
Charles Sawyers (Co-Chair)
Memorial Sloan Kettering Cancer Center, New York, United States
Gil Alterovitz
Harvard Medical School, Boston, United States
Kym Boycott
Children’s Hospital of Eastern Ontario Research Institute, Canada
Paul Burton
University of Bristol, Bristol, United Kingdom
Mark Lawler
Queen’s University Belfast, Belfast, United Kingdom
John Mattison
Kaiser Permanante, Pasadena , United States
Andrew Morris
University of Dundee, Dundee, United Kingdom
Anthony Philippakis
Genome Bridge LLC, Cambridge, United States
Heidi Rehm
Harvard Medical School, Boston, United States
Dan Roden
Vanderbilt University School of Medicine, Nashville, United States
Lillian Siu
University Health Network, Toronto, Canada
Khee Chee Soo
National Cancer Centre of Singapore, Singapore
Volker Straub
Newcastle University, Newcastle upon Tyne, United Kingdom
Mandate and Scope
Mandate:
How do we represent phenotypic data and link it to genotypic
information?
Scope:
Initial focus on rare genetic disorders and cancer
Widen scope to complex traits and ID with time
Approach:
Avoid re-inventing the wheel
Add value to existing endeavors
Address both research and clinical use scenarios
C
LINICAL
W
ORKING
G
ROUP
Approach
1.
Define the need and potential solutions
2.
Document existing efforts
3.
Assess compatibility of approaches
4.
Identify gaps and propose solutions to fill them
5.
Select demonstration projects
6.
Implement solutions in an R&D environment - and when ready,
integrate into clinical use
7.
Ensure sustainability of the solution and adapt as need
evolves
8.
Communicate globally - white papers, networks, international
forums
C
LINICAL
W
ORKING
G
ROUP
Activities To Date
Learning Tour
Phenotype Ontology
PhenoTips and PhenomeCentral, Recording and Sharing of Detailed Patient Data
Michael Brudno (University of Toronto)
Human Phenotype Ontology
Peter Robinson (Charité Universitätsmedizin Berlin)
GeneMatcher and PhenoDB Tools
Ada Hamosh (Johns Hopkins University School of Medicine)
Data Harmonization
Paul Burton (University of Bristol) and Isabel Fortier (McGill University)
Biomedical Informatics, EHRs and Data Extraction
CancerLinQ: The Future of Cancer Care
Lillian Siu (University Health Network)
Phenotyping in the Electronic Medical Record – Challenges and Opportunities
Dan Roden (Vanderbilt University School of Medicine)
Advances in Health Database Technologies: Use of EHR Data for Clinical Discovery
George Hripcsak (Columbia University)
Rare Disease
Mapping of Existing Tools
Initiative
Ontology
Tool
Database
Matchmaker
Phenotype
Genotype
PhenoTips
X
X
X
DECIPHER
X
X
X
PhenomeCentral
X
X
X
ClinGen
X
X
X
ClinVar
X
X
X
PhenoDB
X
X
X
X
OMIM
X
X
X
HPO
X
X
LDDB
X
X
Gene Matcher
X
X
X
GEM.app
X
X
X
LOVD
X
X
Cafe Variome
X
X
Rare Disease:
Mapping of Existing Initiatives
Group
North America
Europe
International
IRDiRC
CARE for RARE
X
NORD
X
Human Variome Project
X
Genetic Alliance
Orphanet
NIH Office of Rare Disease
Research (ORDR)
RD-Connect
Neuromics
X
ICCG
Example of Approach
Phenotype Ontology
1. Define Need Physicians, researchers and clinical laboratories are in need of systems to enable standardized and structured phenotypic data to be
collected from patients. 2. Document Existing
Efforts
A number of phenotype ontologies exist today including:
● Human Phenotype Ontology (HPO) created by Peter Robinson is the most comprehensive ontology
● PhenoDB created by Ada Hamosh for the Centers for Mendelian Genomics (intended for quick and simple use by a physician)
● London Dysmorphology
● SNOMED (Coding system for electronic health records and other medical data systems)
● UMLS
● MedGen (NCBI ontology)
In addition, several databases exist that catalog diseases including OMIM and Orphanet.
Several tools exist to support data collection using these ontologies including PhenoDB and PhenoTips 3. Assess Compatibility
if Multiple Approaches Exist
Given different uses, it is unlikely that a single ontology could support all needs. Instead, recent efforts have focused on identifying the most commonly used terms and mapping these across all ontologies to support interoperability. This effort, called ICHPT, is being led by Ada Hamosh and Peter Robinson and resulted in the identification of 2300 such terms.
4. Identify Gaps and Propose Solutions
● Clinical areas within these ontologies for which expansion of terms are needed?
o
Work with clinical domain experts to identify gaps and propose new terms to fill them● Are there barriers to implementation within the medical and research environments?
o
Slowness of EHR system developmento
Need to support IT integration of tools into clinical laboratory and research systems5. Implementation into Clinical and Research Use
Provide phenotyping tools that can operate as standalone systems as well as integrated into other systems such as electronic health records.
6. Communicate Communicate the availability of tools and solutions to the broader community through publications, talks and websites.
7. Sustain ● Need to maintain and evolve ontologies to meet community needs.
● Need to update tools as ontologies evolve.
1. Define Need
Physicians, researchers and clinical laboratories are in need of systems to enable
standardized and structured phenotypic data to be collected from patients.
Example of Approach
Phenotype Ontology
1. Define Need Physicians, researchers and clinical laboratories are in need of systems to enable standardized and structured phenotypic data to be
collected from patients. 2. Document Existing
Efforts
A number of phenotype ontologies exist today including:
● Human Phenotype Ontology (HPO) created by Peter Robinson is the most comprehensive ontology
● PhenoDB created by Ada Hamosh for the Centers for Mendelian Genomics (intended for quick and simple use by a physician)
● London Dysmorphology
● SNOMED (Coding system for electronic health records and other medical data systems)
● UMLS
● MedGen (NCBI ontology)
In addition, several databases exist that catalog diseases including OMIM and Orphanet.
Several tools exist to support data collection using these ontologies including PhenoDB and PhenoTips 3. Assess Compatibility
if Multiple Approaches Exist
Given different uses, it is unlikely that a single ontology could support all needs. Instead, recent efforts have focused on identifying the most commonly used terms and mapping these across all ontologies to support interoperability. This effort, called ICHPT, is being led by Ada Hamosh and Peter Robinson and resulted in the identification of 2300 such terms.
4. Identify Gaps and Propose Solutions
● Clinical areas within these ontologies for which expansion of terms are needed?
o
Work with clinical domain experts to identify gaps and propose new terms to fill them● Are there barriers to implementation within the medical and research environments?
o
Slowness of EHR system developmento
Need to support IT integration of tools into clinical laboratory and research systems5. Implementation into Clinical and Research Use
Provide phenotyping tools that can operate as standalone systems as well as integrated into other systems such as electronic health records.
6. Communicate Communicate the availability of tools and solutions to the broader community through publications, talks and websites.
7. Sustain ● Need to maintain and evolve ontologies to meet community needs.
● Need to update tools as ontologies evolve.
2. Document Existing Efforts
A number of phenotype ontologies exist today including:
Human Phenotype Ontology (HPO) created by Peter Robinson is the most comprehensive
ontology
PhenoDB created by Ada Hamosh for the Centers for Mendelian Genomics (intended for
quick and simple use by a physician)
London Dysmorphology
SNOMED (Coding system for electronic health records and other medical data systems)
UMLS
MedGen (NCBI ontology)
Several databases exist that catalog diseases including OMIM and Orphanet.
Example of Approach
Phenotype Ontology
1. Define Need Physicians, researchers and clinical laboratories are in need of systems to enable standardized and structured phenotypic data to be
collected from patients. 2. Document Existing
Efforts
A number of phenotype ontologies exist today including:
● Human Phenotype Ontology (HPO) created by Peter Robinson is the most comprehensive ontology
● PhenoDB created by Ada Hamosh for the Centers for Mendelian Genomics (intended for quick and simple use by a physician)
● London Dysmorphology
● SNOMED (Coding system for electronic health records and other medical data systems)
● UMLS
● MedGen (NCBI ontology)
In addition, several databases exist that catalog diseases including OMIM and Orphanet.
Several tools exist to support data collection using these ontologies including PhenoDB and PhenoTips 3. Assess Compatibility
if Multiple Approaches Exist
Given different uses, it is unlikely that a single ontology could support all needs. Instead, recent efforts have focused on identifying the most commonly used terms and mapping these across all ontologies to support interoperability. This effort, called ICHPT, is being led by Ada Hamosh and Peter Robinson and resulted in the identification of 2300 such terms.
4. Identify Gaps and Propose Solutions
● Clinical areas within these ontologies for which expansion of terms are needed?
o
Work with clinical domain experts to identify gaps and propose new terms to fill them● Are there barriers to implementation within the medical and research environments?
o
Slowness of EHR system developmento
Need to support IT integration of tools into clinical laboratory and research systems5. Implementation into Clinical and Research Use
Provide phenotyping tools that can operate as standalone systems as well as integrated into other systems such as electronic health records.
6. Communicate Communicate the availability of tools and solutions to the broader community through publications, talks and websites.
7. Sustain ● Need to maintain and evolve ontologies to meet community needs.
● Need to update tools as ontologies evolve.
3. Assess Compatibility if Multiple Approaches Exist
Given different uses, it is unlikely that a single ontology could support all needs.
Instead, recent efforts have focused on identifying the most commonly used terms
and mapping these across all ontologies to support interoperability. This effort,
called ICHPT, is being led by Ada Hamosh and Peter Robinson and resulted in the
identification of 2300 such terms.
Example of Approach
Phenotype Ontology
1. Define Need Physicians, researchers and clinical laboratories are in need of systems to enable standardized and structured phenotypic data to be
collected from patients. 2. Document Existing
Efforts
A number of phenotype ontologies exist today including:
● Human Phenotype Ontology (HPO) created by Peter Robinson is the most comprehensive ontology
● PhenoDB created by Ada Hamosh for the Centers for Mendelian Genomics (intended for quick and simple use by a physician)
● London Dysmorphology
● SNOMED (Coding system for electronic health records and other medical data systems)
● UMLS
● MedGen (NCBI ontology)
In addition, several databases exist that catalog diseases including OMIM and Orphanet.
Several tools exist to support data collection using these ontologies including PhenoDB and PhenoTips 3. Assess Compatibility
if Multiple Approaches Exist
Given different uses, it is unlikely that a single ontology could support all needs. Instead, recent efforts have focused on identifying the most commonly used terms and mapping these across all ontologies to support interoperability. This effort, called ICHPT, is being led by Ada Hamosh and Peter Robinson and resulted in the identification of 2300 such terms.
4. Identify Gaps and Propose Solutions
● Clinical areas within these ontologies for which expansion of terms are needed?
o
Work with clinical domain experts to identify gaps and propose new terms to fill them● Are there barriers to implementation within the medical and research environments?
o
Slowness of EHR system developmento
Need to support IT integration of tools into clinical laboratory and research systems5. Implementation into Clinical and Research Use
Provide phenotyping tools that can operate as standalone systems as well as integrated into other systems such as electronic health records.
6. Communicate Communicate the availability of tools and solutions to the broader community through publications, talks and websites.
7. Sustain ● Need to maintain and evolve ontologies to meet community needs.
● Need to update tools as ontologies evolve.
4. Identify Gaps and Propose Solutions
Clinical areas within these ontologies for which expansion of terms are needed?
Work with clinical domain experts to identify gaps and propose new terms to fill them
Barriers to implementation within the medical and research environments.
Slowness of EHR system development
Example of Approach
Phenotype Ontology
1. Define Need Physicians, researchers and clinical laboratories are in need of systems to enable standardized and structured phenotypic data to be
collected from patients. 2. Document Existing
Efforts
A number of phenotype ontologies exist today including:
● Human Phenotype Ontology (HPO) created by Peter Robinson is the most comprehensive ontology
● PhenoDB created by Ada Hamosh for the Centers for Mendelian Genomics (intended for quick and simple use by a physician)
● London Dysmorphology
● SNOMED (Coding system for electronic health records and other medical data systems)
● UMLS
● MedGen (NCBI ontology)
In addition, several databases exist that catalog diseases including OMIM and Orphanet.
Several tools exist to support data collection using these ontologies including PhenoDB and PhenoTips 3. Assess Compatibility
if Multiple Approaches Exist
Given different uses, it is unlikely that a single ontology could support all needs. Instead, recent efforts have focused on identifying the most commonly used terms and mapping these across all ontologies to support interoperability. This effort, called ICHPT, is being led by Ada Hamosh and Peter Robinson and resulted in the identification of 2300 such terms.
4. Identify Gaps and Propose Solutions
● Clinical areas within these ontologies for which expansion of terms are needed?
o
Work with clinical domain experts to identify gaps and propose new terms to fill them● Are there barriers to implementation within the medical and research environments?
o
Slowness of EHR system developmento
Need to support IT integration of tools into clinical laboratory and research systems5. Implementation into Clinical and Research Use
Provide phenotyping tools that can operate as standalone systems as well as integrated into other systems such as electronic health records.
6. Communicate Communicate the availability of tools and solutions to the broader community through publications, talks and websites.
7. Sustain ● Need to maintain and evolve ontologies to meet community needs.
● Need to update tools as ontologies evolve.
5. Implementation into Clinical and Research Use
Provide phenotyping tools that can operate as standalone systems as well as integrated
into other systems such as electronic health records.
6. Communicate
Communicate the availability of tools and solutions to the broader community through
publications, talks and websites.
7. Sustain
Need to maintain and evolve ontologies to meet community needs.
Need to update tools as ontologies evolve.
Phenotype Ontology
PhenoTips
Source:
http://phenotips.org/FeaturePreview
PhenoTips
Source: http://phenotips.org/FeaturePreview
Configurable Phenotype Checklist
PhenoTips
Source: http://phenotips.org/FeaturePreview
Measurement Recording and Interpretation
PhenoTips
Source: http://phenotips.org/FeaturePreview
Pedigree Drawing
Matchmaker
1. Define Need The existence of extremely rare phenotypes and genetic variation results in the need for systems to support the algorithmic
matching of unsolved genomic cases based upon overlapping phenotypic and genotypic data. Matching can be simplistic using high level phenotypes and candidate genes or comprehensive (deep phenotypes and full vcfs)
2. Document Existing Efforts ● Gene Matcher within PhenoDB created by Ada Hamosh
● PhenomeCentral created by Michael Brudno
● DECIPHER created by Matt Hurles and Helen Firth
● GEM.app creatde by Stephen Zuchner
● LOVD created by Johan den Dunnen
● Café Variome creatde by the Gen2Phen project
3. Assess Compatibility if Multiple Approaches Exist
● Systems are currently not interconnected.
● Near term proposal is to interface local systems with each system acting as portal to the others.
● Long term solution may involve a centralized site.
4. Identify Gaps and Propose Solutions
● Need to assess data models for each system and agree upon a common format for data exchange
● Need to develop algorithms to enable phenotypic and genotypic matching with scores for degree of match
5. Implementation into Clinical and Research Use
Work with other sequencing sites to encourage use of one of the interfaced systems
6. Communicate Communicate the availability of matching and storage systems to the broader community of clinical labs, physicians and
researchers through publications, talks and websites.
7. Sustain ● Need to determine which resources are needed to sustain the endeavor and how will the resources be provided.
● Need buy-in from data suppliers and data users.
1. Define Need
The existence of extremely rare phenotypes and genetic variation results in the need for
systems to support the algorithmic matching of unsolved genomic cases based upon
overlapping phenotypic and genotypic data.
Matching can be simplistic using high level phenotypes and candidate genes or
comprehensive (deep phenotypes and full vcfs)
Data Harmonization
Need:
Analyses regularly demand large numbers of individuals each measured carefully for a variety of
phenotypes. Questions cannot be answered using any one single study and it is often essential that
data can be brought together across several sources
Key Issues:
Consideration of ethic-legal frameworks that apply to different studies
There are numerous approaches to harmonization – each with pros and cons. However, whichever
approach is decided, it must be consistently adopted for each variable across all studies
Once-data have been epidemiologically harmonized, one of the key challenges is how to co-analyze
the data
Harmonization is labor intensive and context specific and therefore entails substantial additional work
each time a new project is proposed. Therefore rules and algorithms need to be reported
Need for support from funders and journals to require data harmonization
Existing Efforts
COGENE
GenomEUtwin
P
3G
PHOEBE
DataSHIELD
BioShaRE-eu
BBMRI
OBiBa
Maelstrom
Research
DataSHaPER
PHENX Toolkit
CLOSER
Data Extraction from EHRs
Need:
Validated methods to extract phenotypic information from EHRs with varying
architectures
Key Issues:
A larger catalog of validated phenotypes
Validation of automated methods for phenotype development and data extraction
from EHRs
Existing Efforts
eMERGE
HL7
Biobank UK
Kadoorie
biobank
Kaiser Northern
California
Rare Disease
IRDiRC Road Map
Courtesy of Kym Boycott
Avoid reinventing the wheel
Add value to existing endeavours
Convene-Collaborate-Catalyse
Globally
Cancer
1. Clinical Implementation of Next Generation Sequencing:
Defining actionability
(druggable, predictive, prognostic)
Functional significance of variants detected
(e.g. known in same tumor type, known in another tumor type, uncertain)
Actionable Gene Panels for Cancer: consensus
2. Linking phenotyping data to genomic data:
Shared goals with rare disease initiatives
3. Global “Matchmaking” for rare tumors/ mutations:
Sharing information for potential clinical trials
Cancer
Mapping of Current Initiatives
Institution, Consortium or Region Trial or Program Name Profiling Platform/s or Technique/s Genes and or Mutations Cancer(s) Archival or Fresh Tumour Ref And/or Trial Identifier
Cancer Research UK Stratified Medicine
Programme
PCR 9 genes Melanoma, NSCLC, CRC, Breast, Prostate, Ovarian cancer
Archival 121 FISH 3 genes
Dana Farber Cancer
Institute PROFILE Sequenom
OncoMap 41 genes, 471
mutations All solid tumours Archival
122
Institut Curie, Institut
National du Cancer SHIVA
Ion Torrent PGM Ampliseq 46 genes
All solid tumours Fresh biopsy NCT01771458123
Cytoscan HD 29 genes
Institut Gustav Roussy (non-pediatric trials)
MOSCATO aCGH PCR
Not Applicable 96 mutations
Solid tumour phase I
patients Fresh Biopsy NCT01566019
75
SAFIR01 aCGH PCR
Not Applicable
2 genes Breast cancer Fresh Biopsy NCT01414933
124 MSN PCR FISH Seqcan 30 genes 5 genes Melanoma, SCLC, NSCLC Fresh Biopsy 125 Massachusetts General
Hospital NS SNaP Shot 14 genes, >50 mutations
NSCLC, CRC, Melanoma,
Breast cancer Archival
126,127
MD Anderson Cancer Center
T9 Program Sequenom 40+ genes All solid tumours Archival 128 IMPACT
PCR 10 genes
All solid tumours Archival 73
NCT00851032 FISH 1 gene
Clearing House
Protocol PCR ~100 genes All solid tumours
Archival and or Fresh Biopsy
129
Cancer
Mapping of Current Initiatives
Institution, Consortium or Region Trial or Program Name Profiling Platform/s or Technique/s Genes and or Mutations Cancer(s) Archival or Fresh Tumour Ref And/or Trial Identifier Memorial Sloan-Kettering Cancer Center IMPACT
Illumina HiSeq 275 genes (Research
assays) All solid tumours Archival NCT01775072 Sequenom or MiSeq NS (Clinical assays)
Netherlands
Centre for Personalized Cancer Treatment
Ion Torrent PGM ~150 genes
Solid tumours Fresh Biopsy 130 SOLiD 5500xl >2000 genes
Norwegian Cancer Genomics
Consortium
Nationwide
program NS Whole Exome
9 tumour types both solid and haematopoietic Archival and or Fresh Biopsy 131 Princess Margaret
Cancer Centre IMPACT
MiSeq TSACP 48 genes, >700
mutations Selected solid
tumours Archival NCT01505400
60
Sequenom Customized panel 23 genes, 279 mutations Vall D’Hebron
Institute of Oncology NS
Sequenom OncoCarta 19 genes, 238 mutations
Breast cancer, Solid tumour phase I patients
Archival 72,132 Illumina GAIIx NS
Vanderbilt-Ingram
Cancer Center PCMI SNaPshot 6-8 genes, >40 mutations
Melanoma, NSCLC,
CRC, Breast cancer Archival
133
WIN Consortium WINTHER study NGS NS Solid tumours
Fresh Biopsy (Tumour and Matched Normal)
NCT0185629683
CWG Next Steps
Clinical Working Group
Tomorrow’s Goals
Meeting Objectives:
To identify the best driver/demonstration projects for the CWG
To expand on the work currently being done in Rare Disease and Cancer
To determine CWG next steps (1 and 3 year plan) and timeline
Cancer Topics:
Clinical Implementation:
Next generation sequencing: overcoming barriers to implementation in the clinic
Actionable gene panels for cancer: achieving a consensus
Phenotyping and clinical data: agreeing on the right solutions and linking with genomic data
Global “matchmaking” for rare tumors/ mutations
Rare Disease Topics:
Phenotype ontology
Electronic Health Record: linking ontologies in EHRs
Patient registries