The Matchmaker Exchange

The Matchmaker Exchange:
Connecting Matchmakers to Accelerate Gene Discovery
Heidi Rehm, PhD, FACMG
Associate Professor of Pathology, Brigham & Women’s Hospital and Harvard Medical School
Director, Partners Laboratory for Molecular Medicine
Clinical Director, Broad Institute Clinical Research Sequencing Platform
On behalf of the MME Working Group
Goal: To identify causal genes for rare disease by matching
phenotypes and genotypes across a federated network of
genomic databases.
Patient #1
Clinical Geneticist #1
Patient #2
Clinical Geneticist #2
Notification
of
Match
Phenotypic
Data
Feature 1
Feature 2
Feature 3
Feature 4
Feature 5
Genotypic Data
Gene A
Gene B
Gene C
Gene D
Gene E
Gene F
Genomic
Matchmaker
Genotypic
Data
Gene D
Gene G
Gene H
Phenotypic
Data
Feature 1
Feature 3
Feature 4
Feature 5
Feature 6
Courtesy of Joel Krier
The state of affairs as of the initial meeting in October 2013, ASHG.
Gene
Matcher
PEER
DECIPHER
Genome
Connect
Monarch
LOVD
Multiple
Matchmaker
disconnected
Exchange
projects
Café
Variome
Undiag.
Diseases
Program
ClinGen
Gene
GEM.app
Phenome
Central
Yenta
Making a Match
Linking multiple databases is not without
its challenges…
The Real Story
Gene
Matcher
Phenome
Central
Exchanging Data
API v 1.0 0.1
DECIPHER
Courtesy Ben Hutton, DECIPHER
Gene
Gene
Matcher
Diseases
PEER
Gene and Phenotype (HPO)
DECIPHER
Variants
Disease and Variants
Genome
Connect
Model Organisms
Monarch
LOVD
Multiple
Matchmaker
disconnected
Exchange
projects
Café
Variome
Gene Undiag.
and Phenotype (HPO)
Variant and Disease
ClinGen
Disease and VCFs
Variants and Phenotype
Diseases
Program
Gene
GEM.app
Phenome
Central
Yenta
Gene and Phenotype (HPO)
Steps to Build the MME
• Defined MME API – aided by GA4GH Data WG
• Created consent policy – aided by GA4GH REWG
(see Appendix)
• Determined requirements to be a MME service
(see Appendix)
• Developed user agreement to use the MME (see
Appendix)
• Security requirements - work in progress with
GA4GH Security WG
PhenomeCentral
www.phenomecentral.org
Dr. Mike Brudno
U of T
Step 1: Add Patient via PhenoTips
Courtesy of Mike Brudno
API: Query by Example
Request: a (real) patient profile, with phenotype
and/or gene/genotype data
Response: a list of “similar” patient profiles
Note: Matching algorithms under refinement
Step 2: See Patients Similar to Yours
Courtesy of Mike Brudno
Step 3: Contact Submitter of Other Dataset
Courtesy of Mike Brudno
MME Pilot
GeneMatcher
• Designed to connect clinicians/researchers
with the same candidate gene (e.g. from a
patient, animal models, etc.)
• Only a human gene symbol or ID is required.
• Can also match on OMIM phenotype
numbers.
• Matching on phenotypic features is in
development.
As of 6/1/2015:
2178 genes submitted by
486 submitters from
38 countries
307 Matches!
Courtesy Ada Hamosh, François Schiettecatte, Nara Sobreira
Growth in Number of Genes & Matches
in GeneMatcher
2500
Number
2000
1500
1000
500
0
Courtesy Ada Hamosh, François Schiettecatte, Nara Sobreira
Genes
Matches
16
MME Pilot
• Test set of 50 solved cases into both databases
– All successfully identified
• Tested using 45 unsolved cases with flagged
candidates in PhenomeCentral
– 10 gene matches
• 6 false positives (discordant phenotypes)
• 2 unresolved
• 2 potentially significant
Currently Connected MME Services
Special Issue of Human Mutation
o Guest Editors: Ada Hamosh, Kym Boycott, Heidi Rehm
o Published October (timed with MME session at ASHG)
o 16 papers
• MME Overview
• MME API
• Mathematics of Matching Algorithms
• Papers by MME services (PhenomeCentral, DECIPHER,
GeneMatcher, Café Variome, Monarch, UDN,
GENESIS/GEM.app, GenomeConnect, Patient Matching)
• Cases solved through matching (3 papers)
How to Participate
Users can either connect a database to the MME
via the API
or
Choose an existing MME service
Guiding Use by the Community
These tables help define the database location, content and approaches for each
MME service to help guide the user in choosing a MME service for case deposition.
Phenotypes
Matchmaker
Exchange Site
Genotype
Chromosome
Structured Non-Human
Variants VCF Files
Gene Name
coordinates
Phenotypes
Models
Diagnosis
term
Diagnosis
code
PhenomeCentral
(Canada)
√
√
√
GeneMatcher
(USA)
√
√
√
DECIPHER
(United Kingdom)
√
Matchmaker
Exchange Site
√
√
Parameters Used for Matching
Gene
Diagnosis
PhenomeCentral
√
√
GeneMatcher
√
√
DECIPHER
√
Candidates
Phenotype
Match Score
Features
√
√
√
√
√
√
√
√
√
√
√
√
Evidence
Flagged
for Gene
Gene
Candidate
Candidates
s
√
√
√
√
www.matchmakerexchange.org
Future Directions
• Support for hypothesis-free matching in the
absence of an identified candidate gene
• Designate more case data as freely
accessible for any type of analysis (defining
phenotypes, meta-data analysis, algorithm
training, etc)
• Patient-initiated matching
Acknowledgements
S Balasubramanian
Mike Bamshad
Sergio Beltran Agullo
Jonathan Berg
Kym Boycott
Anthony Brookes
Catherine Brownstein
Michael Brudno
Han Brunner
Orion Buske
Deanna Church
Raymond Dalgliesh
Victor de la Torre
Andrew Devereau
Johan den Dunnen
Sergiu Dumitriu
Helen Firth
Paul Flicek
Jan Friedman
Richard Gibbs
Marta Girdea
Robert Green
Ada Hamosh
Melissa Haendel
Ingrid Holm
Matt Hurles
Ben Hutton
Ekta Khurana
Sebastian Kohler
Joel Krier
Owen Lancaster
Melissa Landrum
Farrah Ladha
Paul Lasko
Rick Lifton
Daniel MacArthur
Alex MacKenzie
Danielle Azzariti
Aleksander Milosavljevic
Chris Mungall
Debbie Nickerson
Woong-Yan Park
Justin Paschall
Anthony Philippakis
Heidi Rehm
Peter Robinson
Gary Saunders
Francois Schiettecatte
Rolf Sijmons
Nara Sobreira
Jawahar Swaminathan
Morris Swertz
Peter Taschner
Sharon Terry
Rachel Thompson
Stephan Zuchner
Appendix
Launching the MME
1. Goals of MME Defined – At the 2013 ASHG Conference, representatives of genomic discovery platforms met and agreed on
a common goal to develop a federated network of databases to support novel gene discovery and inform the clinical
interpretation of genomic variation.
2. Version 0.1 MME API Developed – In 2014, members of PhenomeCentral and GeneMatcher collaborated to develop a
common API. The group engaged technical input from the wider MME services and the Data Working Group of the GA4GH
to refine the API and conform to evolving genomic standards. The API was posted on GitHub for community exchange of
ideas.
3. MME Core Policies Developed – Over the span of one year, core policies to support the MME were developed and finalized
in January 2015 at the Miami MME meeting. These policies are described above and include a tiered consent framework, a
memorandum of understanding for MME services, a user agreement and a steering committee to oversee the program.
4. Matchmaker Exchange Website Launched – a website was launched to begin engagement of the outside community. The
website can be found at www.matchmakerexchange.org.
5. Matching Algorithm Principles Defined – It was agreed that each matchmaker service would support its own matching
algorithms, in part to deal with the practical aspects of different data schemas, database content and notification
management on behalf of existing users. However, matchmaker services continue to collaborate and share practices using
GitHub as a forum for exchange of ideas and algorithms.
6. API Test Phase – Beginning in 2014, three matchmaker services (PhenomeCentral, GeneMatcher and DECIPHER)
implemented the API in three point-to-point launches in which the entire content of each database was queried against the
content of the other databases to evaluate the code-base for the API and matching algorithms. API keys are exchanged
using a standard Apache-based protocol that promotes data security. Problems were documented on a GitHub issues
tracker and resolved.
7. MME Test Dataset Developed – To improve the ease with which new matchmaker services can be incorporated into the
MME, a test dataset has been developed and resides on GitHub to allow evaluation of API and matching algorithm
implementations.
8. User Interface to Support Queries – User interfaces at individual MME services are being built to support user-initiated
queries of content across the MME. The status of user interfaces for each MME service can be found on the MME website.
API v1 Data Fields
ID (Mandatory) - The internal identifier (obfuscated or not) that can be used by the originating system to
reference the patient data.
Label (Optional) - A name/identifier assigned by the user which can be used to reference the patient in a
recognizable manner (in an email for example); it should not contain any personally identifiable information.
Query type (Optional)
–
Accepted values: once or periodic
Submitter (Mandatory) - Consists of contact information of the person that submitted the search
Gender (Optional)
Age of onset (Optional)
Mode of inheritance (Optional)
Disorders (Optional) - A list of OMIM (MIM:######) or OrphaNet (ORPHA#####) identifiers, can be empty
Features – It is mandatory to have at least clinical features or gene(s ); having both is preferred
–
HPO terms for clinical features
–
gene name(s)
MME Informed Consent Policy
The MME worked closely with the GA4GH's Regulatory and Ethics Working Group and Consent Task
Team on developing a proposal for informed consent for data sharing in the context of genomic
matchmaking within MME. We have distinguished two levels of matchmaking and different consent
requirements based on the data shared and the probability of re-identifying the patient:
Level 1 - No additional consent required - This level of matchmaking involves a data requester
querying on a broad phenotype description or disease name using standardized terms or codes
(Human Phenotype Ontology (HPO), OMIM, Orphanet) and/or candidate gene names +/- variant
type. This level of sharing is consistent with current clinical practice with low risk of possible reidentification and therefore specific patient consent for this activity is not required.
Level 2 - Consent required - This level of matchmaking involves a data requester querying on a
unique or sensitive phenotype description and/or sequence level and related information, such as
defined variants and/or genomic datasets. This level of sharing requires consent from the patient.
If the patient had previously consented to data being shared in an open or registered access
database whose declared purpose involves data sharing for purposes consistent with those of this
matchmaking, no additional consent is required.
The MME service in which data is deposited is responsible for ensuring patient data used in
matchmaking is consented appropriately.
MME Service Requirements
To become a MME service, each new site must achieve the following:
1. Require users to deposit case data to undertake a federated query across the MME service
providers
2. Establish a minimum of two point-to-point API connections to other MME services
3. Contain content that is considered by the MME steering committee to be useful for matching,
including the flagging of, or ability to prioritize, candidate genes
4. Successfully implement matching algorithms using test data
5. During user queries, enable dual notification of data requester (i.e. the querier) and prior data
depositor (i.e. the queried) including sharing the identities and contact information for each
6. For each database to which a MME service is connected by an API, the connected database’s
disclaimers should be posted on the MME service’s website and displayed with query results.
Disclaimers can be found on GitHub (https://github.com/MatchmakerExchange/mmeapis/tree/master/disclaimers)
7. Store queries sent and received between MME sites only for the purpose of auditing, defining
query statistics, and following up queries to understand rates of validated gene discovery
8. Attest to database security requirements as defined by the GA4GH Security WG (forthcoming)
9. Advance the goals of the MME project through active participation in meetings and
conference calls including defining a representative for the MME steering committee
MME End User Agreement
To use the MME, each data querier agrees to the following:
1. To make no attempt to identify individual patients in any MME
database
2. To enable all cases submitted for querying to be stored in the
query-initiating database for future matching
3. To obtain permission from the source of the matching data
before publishing or presenting the results of queries
4. To acknowledge the MME, and the specific MME service that
supported any discoveries in publications, as appropriate