1 of 10

How can we provide

computable evidence for a diagnosis?

What is the evidence that a disease exists?

2 of 10

xxxxx

Understanding equivalency: What is the evidence that these 11 records for “Ehlers Danlos Syndrome” are related?

Narrow synonym? Broad? Exact? Child? Parent?

Relevant papers:

Bayesian models like k-BOOM can help. Mungall, et al. doi:10.1101/048843

bit.ly/xref-wildwest

A Census of Disease Ontologies

Haendel, et al. https://doi.org/10.1146/annurev-biodatasci-080917-013459

Many disease ontologies and terminologies contain disease terminology and classifications. Key features of these terminologies are a 1) label: a unique, primary label for the term; 2) definition: a textual definition of the term; 3) synonyms: synonyms that are narrow (more specific), broad (more general), or exact; and 4) child terms: a ‘parent-child’ classification, where terms are arranged in a hierarchy, and more general terms are classified as ‘parents’ or super-classes of more specific terms (children or subclasses).

This table lists a number of disease terminologies and ontologies, and example IDs and labels corresponding to terms from the respective terminology. It is noted if the term contains a text definition, the number of synonyms and the number of child terms. The boxed area of the table represents outgoing mappings - where an ontology/terminology cross-references other classes in another terminology if they are equivalent terms, in order to provide association between one data item and another. For example, Mondo:0020066 cross references several terms, including one from Orphanet, one from MESH, etc. Mondo also has 475 disease-phenotype associations (ie phenotypes that are typically seen in patients with this disease).

The Distinct row indicates the number of unique data points for each column, e.g., there are 5 distinct labels, and 59 distinct synonyms.

Abbreviations: AOD, Alcohol and Other Drug Thesaurus; COSTAR, Computer Stored Ambulatory Record; CTEP, Cancer Therapy Evaluation Program; DO, Human Disease Ontology; The Genetic and Rare Diseases (GARD) Information Center; ICD-9/-10, International Classification of Diseases, Ninth Revision/Tenth Revision; KEGG, Kyoto Encyclopedia of Genes and Genomes; MedDRA, Medical Dictionary for Regulatory Activities; MedGen, National Center for Biotechnology Information Medical Genetics; MeSH, Medical Subject Headings; Mondo, Monarch Disease Ontology; NCI, National Cancer Institute Thesaurus; OMIM, Online Mendelian Inheritance in Man; PDQ, Physician Data Query; SNOMED, Systematized Nomenclature of Medicine; UMLS, Unified Medical Language System.

3 of 10

Harmonizing diseases: Mondo disease ontology

We needed:

Disease categories spanning multiple categories
A systematic way of relating the concepts to one another

https://mondo.monarchinitiative.org/

CANCER

COMPLEX

INFECTIOUS

MENDELIAN

RARE

...

4 of 10

Standards proliferation: how do you know you need a new one?

xkcd.com/927

5 of 10

Standards proliferation: how do you know you need a new one?

For Diseases:

SITUATION:

THERE ARE

15*14=210

SETS OF

MAPPINGS.

6 of 10

Evidence-based merging of equivalent classes

...

MEDIC

MESH

DC

DO

EFO

GARD

NCIT

Orphanet

kBOOM

Bayesian

OWL

Ontology

Merging

Logical +

Probabilistic

Inference

Curated Equivalence

Relations

evaluate

iterative curator-assisted generation

curate

feedback

OMIM

To computationally utilize our collective knowledge sources for diagnostics and to reveal underlying mechanisms of diseases, we need to understand which terms are equivalent across different resources. This will allow the integration of associated information, such as treatments, genetics, phenotypes, etc. To that end, we created the Mondo Disease Ontology (https://monarch-initiative.github.io/mondo/), which provides a logic-based structure that unifies multiple disease resources. Mondo is created by a combination of a Bayesian approach to ontology merging (k-BOOM) and iterative expert curation. Mondo integrates several disease ontologies, some examples are listed in this slide. For a complete set of sources, see: https://mondo.monarchinitiative.org/pages/sources/.Mondo provides equivalence mappings to other disease resources, and annotates each mapping using strict semantics, so that precisely equivalent diseases can be distinguished from those that are merely closely related - allowing improved computational integration of associated data.

7 of 10

From Practice-based Evidence�to Evidence-based Practice

Clinical

Databases

Registries

et al.

Clinical

Guidelines

Expert

Systems

Data

Inference

Knowledge

Management

Decision

support

Terminologies and data models provide the consistency and comparability

essential for a Learning Health System

Patient

Encounters

Medical

Knowledge

Terminologies

Data models

8 of 10

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

9 of 10

If rare diseases are not counted,

rare disease patients will not count

10,577 rare disease concepts

~50% higher than most

prior estimates

In an effort to quantify all the rare diseases, an analysis, where we generated a TSV containing 21633 Mondo disease terms, each with annotations that signifies whether the term is a rare disease term and whether that term is a leaf node in the ontology. The horizontal bars on the left represent the total number of disease concepts in each of five disease resources. The vertical bars count disease “leaf” concepts and are sorted in descending order. The dots highlight the visual overlap between the five sources within the harmonized Mondo resource. Orphanet and OMIM have the highest number of “unique” rare disease concepts, with the next one showing overlap between GARD, Orphanet and OMIM for 982 unique rare disease leaf terms. Only 333 rare diseases match across all five resources (the bar with 5 black dots). We consider that the combination of all five resources provides a much better approximation of the number of rare diseases than historical values. For more details, see: https://mondo.monarchinitiative.org/pages/analysis/ and https://zenodo.org/record/3478576#.XaC8T5NKgl5

10 of 10

Semantically anchoring ICD, LOINC, FHIR with Mondo & HPO (more logically interoperable ontologies) can improve interoperability

MONDO, and HPO can provide the semantic anchoring to ICD-11, LOINC, and FHIR to improve both clinical and research data interoperability.