1 of 22

Melissa Haendel, PhD

Modeling and making computable the complex interplay between environmental and genetic causes and phenotypic outcomes.

2 of 22

Orphan Drug Act of 1983

https://rarediseases.org/orphan-drug-act-resolution-introduced-in-congress/

https://globalgenes.org/rare-facts/

3 of 22

Why the number of rare diseases is hard to determine (and is not 7000)

We don’t have the same criteria for “rare” around the world:

1983: From the Orphan Drug Act: A rare disease affects fewer than 200,000 people
2000: European Union considers a disease to be rare when it affects fewer than 1 in 2,000 people.

We add new diseases all the time, but don’t update the number:

New rare diseases are discovered every week by organizations such as the Undiagnosed Disease Network
The literature, OMIM, etc. abound by new weekly entries
N-of-1s are matched, defining new diseases in systems like the Matchmaker Exchange

We don’t define diseases in the same way

Dozens of terminologies and disease registries exist
These are often not included in clinical terminologies (such as ICD) commonly used in EHRs
Fundamentally, the definition of a rare disease and how to model it computationally has remained more an art than a science

4 of 22

Why do I care about this?

=> Not having clear definitions of rare diseases makes them harder to build diagnostic tools and reveal mechanisms

=> yes identifiers matter!

Identifiers for the 21st century: How to design, provision, and reuse persistent identifiers to maximize utility and impact of life science data

5 of 22

hpo.jax.org

Phenotyping terminology�>14,500 terms

Computational disease models�>190,000 disease-phenotype annotations

Widely adopted in rare disease genomic diagnostic tools

100,000 Genomes Project, SOLVE-RD, NIH-UDP, etc.

Human Phenotype Ontology (HPO)

6 of 22

Fuzzy Phenotype Matching

Perfect Match

Fuzzy Match

No Match

Legend

DOI: 10.1126/scitranslmed.3009262

Not same variant, but same disease and gene, KMT2A.

7 of 22

What is the most clinically useful way to

define and group diseases?

CANCER

COMPLEX

INFECTIOUS

MENDELIAN

RARE

We needed:

Disease concepts spanning multiple categories
A systematic way of relating these concepts

Why not just use mappings?

Many terminologies / ontologies / lists include mappings

These can be used to cross-walk

Problems:

Often mutually inconsistent
N^2 sets of mappings!
Not 1:1 equivalents

mondo.monarchinitiative.org

8 of 22

Where did the definitions come from?

How do they relate?

bit.ly/disease-census

Narrow synonym? Broad? Exact? Child? Parent?

Bayesian models like k-BOOM can help

Mungall

doi:10.1101/048843

9 of 22

Mappings are insufficient

Proliferation of mappings

Too many combinations
Frequently conflicting
Frequently stale

Semantics unclear

Equivalent?
Exact?
Broad/Narrow/Related?
Without precise equivalence mapping, merging is not possible

C1

C2

C3

C4

C6

C5

(N^2)-N sets of mappings (if each source provides their own mappings to all)

10 of 22

Example of problem with cycles

C0442874

Neuropathy

C0031117

Peripheral neuropathy

is-a

https://github.com/NCI-Thesaurus/thesaurus-obo-edition/issues/20

C4731�Neuropathy

C27580�PNS Disorder

C119734�Peripheral Neuropathy�

External Resource

(NCIT)

11 of 22

Different communities annotate different relationships, at different levels of granularity and using different vocabularies

12 of 22

Systematically lump and split

to achieve disease harmony

bit.ly/lump-split

bit.ly/mondo-io

13 of 22

Evidence-based merging of �equivalent disease concepts

...

MESH

DC

DO

EFO

GARD

NCIT

Orphanet

k-BOOM

Bayesian

OWL

Ontology

Merging

Logical +

Probabilistic

Inference

Curated Equivalence

Relations

evaluate

iterative curator-assisted equivalence inference

curate

feedback

OMIM

doi.org/10.1101/048843

MEDIC

14 of 22

15 of 22

Each gold standard disease has a phenotype profile

# Associated Diseases

16 of 22

Overall design principles

https://github.com/monarch-initiative/mondo/tree/master/src/patterns/dosdp-patterns

Dead Simple Ontology Design Patterns (DOSDP)

Pre-made patterns that specify:

label
text definition
synonyms
logical definition

Disease series by gene

pattern_name: disease_series_by_gene

description: >-

This pattern is for diseases that are caused by a single mutation in a single gene, that have gene-based names, such as new disease terms that are requested by ClinGen, like like MED12-related intellectual disability syndrome.

Examples: [MED12-related intellectual disability syndrome](http://purl.obolibrary.org/obo/MONDO_0100000), [TTN-related myopathy](http://purl.obolibrary.org/obo/MONDO_0100175), [MYPN-related myopathy](http://purl.obolibrary.org/obo/MONDO_0015023)

classes:

disease: MONDO:0000001

gene: SO:0001217

relations:

disease has basis in dysfunction of: RO:0004020

vars:

disease: "'disease'"

gene: "'gene'"

name:

text: '%s caused by mutation in %s'

vars:

- disease

- gene

Design patterns allow for consistency amongst terms and consistent, automated classification of the hierarchy

17 of 22

Anatomy of a Mondo term

summary

}

term

info

{

hierarchy

18 of 22

How many rare diseases are there?

NCIT

DOID

GARD

Orphanet

OMIM

Just 5 sources comprise 10,577 unique rare disease concepts

(prior estimates ~7,500)

Only 333 shared disease concepts in all five sources

Many diseases are in only one source

Nature Reviews Drug Discovery (bit.ly/nature-rare-diseases)

Intersection size

5 selected sources

19 of 22

Status and community development

MONDO IDs assigned and tracked for each concept

Use of standard ontology engineering practices

Periodically aligned and synced with existing resources

Released monthly (obo, owl, json)

mondo.monarchinitiative.org

OBO Foundry

obofoundry.org/ontology/mondo

Ontology Lookup Service

ebi.ac.uk/ols/ontologies/mondo

GitHub

github.com/monarch-initiative/mondo

>1300 issues reported

55 releases

Weekly Calls

Fridays, 9am PT/12pm ET

Zoom

Mailing list:

https://groups.google.com/forum/#!forum/mondo-users

Major changes (such as obsoletion candidates or new releases) are shared with the mailing list regularly

Thank you to NHGRI for Phenomics First funding

Where to view Mondo

20 of 22

Users

mondo.monarchinitiative.org

21 of 22

mondo.monarchinitiative.org

https://bit.ly/clingen-mondo

Big thanks to our contributors!

Broad Institute

Samantha Baxter

Andrew Grant

Jessica Hekman

Madeline Hughes

Kate Megquier

Kathy Reinold

Rebecca Siegert

CHOP

Colin Ellis

Allison Heath

Ingo Helbig

Avi Kelman

CoRDS-Sanford

Austin Letcher

ClinGen

Larry Babb

Taylor Bingaman

Marina DiStefano

Jenny Goldstein

Brooke Palus

Heidi Rehm

Erin Riggs

Tam Sneddon

Courtney Thaxton

Matt Wright

EBI

Mélanie Courtot

Simon Jupp

David Osumi-Sutherland

Zoë Pendlington

Paola Roncaglia

GARD

Gioconda Alyea

PJ Brooks

Maria Della Rocca

Janine Lewis

Anne Pariser

Andrea Storm

Monarch Initiative

Melissa Haendel

Leigh Carmody

Shahim Essaid

Nomi Harris

Nico Matentzoglu

Julie McMurry

Moni Munoz-Torres

Peter Robinson

Kent Shefchek

Anne Thessen

Aaron Zhang

NCIt

Gilberto Fragoso

Bron Kisler

NIH NCATS

Alice Chen

Eric Sid

NCBI

Donna Maglott

Johns Hopkins

Christopher Chute

NIH NHGRI

Robert Fullem

Morgan Similuk

NORD

Vanessa Boulanger

OMIM

Joanna Amberger

Ada Hamosh

Orphanet

Marc Hanauer

Annie Olry

Ana Rath

University of Colorado

Tiffany Callahan

22 of 22

http://bit.ly/annual-reviews-ontologies

Wide heterogeneity in:

placement of diseases hierarchically, and therefore meaning,
mapping to other diseases,
the number and typing of synonyms

Assessment of synonyms, hierarchies, and mappings across ontology sources for example diseases EDS and Pancreatic cancer

2018 statistics shown from this manuscript:

Abbreviations: DO, Human Disease Ontology; ICD-10/-11, International Classification of Diseases, Tenth Revision/Eleventh Revision; MedDRA, Medical Dictionary for Regulatory Activities; MeSH, Medical Subject Headings; MonDO, Monarch Disease Ontology; NCIt, National Cancer Institute Thesaurus; OMIM, Online Mendelian Inheritance in Man; SNOMED, Systematized Nomenclature of Medicine; UMLS, Unified Medical Language System; Xref, external reference (cross-reference). aThe Xref and synonym columns summarize the extent and precision of Xrefs and synonyms, respectively, provided by the respective records in the source ontologies. The denominators in these columns refer to the total number of Xrefs or synonyms listed by the source ontology; the numerators refer to the number of these that we manually assessed as equivalent to the concept (not narrower, broader, related, or obsolete). In some cases, something akin to a qualifier (narrower, broader, related, or obsolete) is already provided by the source; in such cases, if we found the qualifier to be accurate by manual review, it was counted toward the numerator. The notation “–’’ means that the denominator was zero (no such term was found). Source data and scripts for these assessments are available at https://github.com/monarch-initiative/ont-review. b“Has definition” refers to whether the record has a textual definition in addition to a label; some of the ontologies also have logical definitions (not shown). cClinical information is provided, but it is not as concise as a conventional definition. dThe scope note is an approximation of a definition. eThe MeSH field entry terms is the closest to synonyms; numbers are especially high due to casing and punctuation variants. fOMIM does not have a record for the group of diseases which comprise Ehlers–Danlos syndrome; these counts reflect those for a specific subtype (Ehlers–Danlos Syndrome, Kyphoscoliotic Type, 1; OMIM:225400).