2 of 25

What is IMPC

Created with BioRender.com

Mendelian diseases

�Hepatic steatosis MP:0002628

IMPC’s objectives:

Mouse KO catalogue
Elucidate function of all mouse protein coding genes
Find new models for human disease

Standardised phenotyping

Initiative to characterise the function of every protein coding gene

Produces resources for external research:

KO mouse lines
Mouse data

3 of 25

QMUL and IMPC

Our role:

Data analysis of the data generated by the IMPC pipeline�
Predict human diseases associated with a gene�
Aid in finding novel disease models�
PhenoDigm algorithm implementation

4 of 25

PhenoDigm and application within IMPC

Phenotype comparisons for DIsease Genes and Models (PhenoDigm)
This quantitative measure can tell us how well our mouse models recapitulate the clinical features of a disease¹
Ontologies: Mammalian Phenotype Ontology (MP) and Human Phenotype Ontology (HPO)
Disease HPO terms aligned with MP terms and produce a percentage similarity score

¹Smedley et al. 2013 Database

PhenoDigm

This phenotypic similarity is computed by several algorithms developed by the Monarch Initiative (Mungall et al. 2017). Firstly, we need to make the two ontologies comparable by correlating every MP term with the corresponding HPO term. Secondly, we compute a score which give us the similarity for each HPO–MP phenotypic match. The PhenoDigm algorithm (Smedley et al. 2013) uses all the individual scores for each HPO–MP association, taking into account the proximity of the two terms in the ontology and the frequency of the phenotype in common from the overall mouse and disease annotations. It finally produces a percentage score by comparing these results to the best possible score (a mouse model perfectly matching the disease). This score allows for the automated identification of mouse models of disease.

5 of 25

Example

Hepatic steatosis

HP:0001397

Fatty liver

disease model

PhenoDigm score

Hepatic steatosis MP:0002628

Phenotype matching using orthologous human disease genes

MP annotations

HPO annotations

Predicted disease model

PD algorithm

6 of 25

Pipeline application overview

PhenoDigm implementation within IMPC

7 of 25

Dissemination methods

IMPC website - disease section

Shiny App (soon to be incorporated to IMPC website)

PhenoDigm scores for pipeline genes
Information available:

Focus on known disease genes
Novel gene discovery

https://diseasemodels.research.its.qmul.ac.uk/

Pilar Cacheiro

8 of 25

Why is this resource important?

10 of 25

IMPC disease models across diverse biological systems

Biological system	Disease Gene	Human Mendelian disease	Relevant Human Phenotype	Overlapping Mouse phenotype
Bone	SCARF2	Van Den Ende-Gupta Syndrome	Long metacarpals	Increased length of long bones
Cardiovascular	LMNA	Cardiomyopathy Dilated 1a	Dilated cardiomyopathy	Increased heart weight
Craniofacial	MSX1	Orofacial Cleft 5	Cleft palate	Cleft palate
Embryo	PSPH	Phosphoserine Phosphatase Deficiency	Intrauterine growth retardation	Abnormal embryo size
Growth/Body size	GHRHR	Isolated Growth Hormone Deficiency, Type Ib	Short stature	Decreased body length
Hearing	SLC52A2	Brown-Vialetto-Van Laere Syndrome 2	Sensorineural hearing impairment	Increased or absent threshold for auditory brainstem response
Hematopoietic	GP9	Bernard-Soulier Syndrome	Thrombocytopenia	Thrombocytopenia
Metabolism	KCNJ11	Diabetes Mellitus, Noninsulin-Dependent	Type II diabetes mellitus	Impaired glucose tolerance
Muscle	COL6A2	Bethlem Myopathy	Distal muscle weakness	Decreased grip strength
Neurological	GOSR2	Epilepsy, Progressive Myoclonic, 6	Difficulty walking	Abnormal gait
Reproductive System	RNF216	Gordon Holmes Syndrome	Infertility	Male infertility
Retina	BBS5	Bardet-Biedl Syndrome 5	Rod-cone dystrophy	Abnormal retina morphology

11 of 25

Bardet-Biedl syndrome-5 and BBS5

•Autosomal recessive ciliopathies caused by mutations in one of 19 genes forming the BBSome

•Bbs5 mice recapitulate retinal dystrophy and obesity as well as exhibiting additional features such as impaired glucose homeostasis

12 of 25

KDELR2 – gene discovery using IMPC data

Osteogenesis imperfecta (MIM:166200)
Neurodevelopmental features
Dysmorphic features

Incompletely penetrant preweaning lethality
Bone structural abnormalities in early adult
Decreased size
Abnormalities in head shape and size
Facial dysmorphology
Abnormal embryonic body wall structure

Efthymiou, S et al., 2021 Am J Med Genet A

13 of 25

Limitations

Certain disorders/physiological systems do not score well

Certain human phenotypes are hard to capture
Some assays in mice do not capture human phenotypes

Abnormal phenotype availability:

Disorders in humans have more annotations compared to mice

Limitations of matching embryo/prenatal phenotypes in humans

15 of 25

PhenoDigm – IMPC pipeline implementation

Download
Build
Owltools
Score
Solr (export a solr core for IMPC website)

16 of 25

Download

Downloads data resources needed by the pipeline

Data dependencies:

Ontologies (hp and mp in .obo format)
IMPC statistical results, phenotype annotations
JAX misc genotype-phenotype associations (MGI)
OMIM disease annotations (morbidmap, mimTitles, mim2gene)
Orphanet disease annotations
HGNC human gene identifiers
HPO genotype-phenotype associations (phenotype.hpoa)
bioMart human-mice gene mappings from Ensembl (currently broken, fix in April/May)

Ontology dependencies: catalog.xml (former catalog-v001.xml):

Ontologies in .owl format:

uPheno
Mp
Hp

17 of 25

Build

Builds and populates a SQL database

Loads data:

Genes & orthologues (from HGNC, MGI, Ensembl)

Ontologies:

Human Phenotype Ontology and Mammalian Phenotype Ontology
Parses elements in hp.obo and mp.obo, terms and synonyms

Diseases:

Known disease genes from OMIM and Orphanet

Models:

IMPC and MGI models

18 of 25

Owtools semantic similarity

Commands to generate support ontologies
owltools –-catalog-xml catalog.xml mp-importer.owl --merge-imports-closure --load-instances Mm-gene-to-phenotype-O.txt --load-labels Mm-gene-labels.txt --merge-support-ontologies -o Mm-all.owl
owltools –-catalog-xml catalog.xml hp-importer.owl --merge-imports-closure --load-instances Hs-disease-to-phenotype-O.txt --load-labels Hs-disease-labels.txt --merge-support-ontologies –o Hs-all.owl

These are very similar to Exomiser too…

19 of 25

Owltools – set up support ontologies

Catalog.xml

mp-importer.owl

Mm-gene-to-phenotype-0.txt

Mm-gene-labels.txt

Owltools

Mm-all.owl

Same process for the hp ontology: hp-importer.owl, Hs-disease-to-phenotype-0, and Hs-disease-labels.txt

Hs-all.owl

This is mp.obo in owl format

20 of 25

Owltools – semantic similarity calculation using support ontologies

3 main files:

Hs-all.owl
Mm-all.owl
mp_hp-align-equiv.owl → mapping

Command:

owltools Hs-all.owl Mm-all.owl mp_hp-align-equiv.owl --merge-support-ontologies --sim-save-phenodigm-class-scores -m 2.5 -x MP,HP -a owltools-cache-hp-mp.txt

Calculates Jaccard Index (simJ) and Information content (IC) of the Least Common subsuming (LCS) phenotype between a pair of concepts

Query term	Match term	simJ	IC	LCS

21 of 25

Score

Query (disease)	Match (gene)	Avg norm	Avg raw	Max norm	Max raw	Query phenotype	Match phenotype

22 of 25

Packaging and deployment

The SQL database → solr core → EBI → New data release

23 of 25

Implementation limitations

Pipeline relies on many separate data sources + overlap with build for Exomiser

Semantic similarity calculation is done using an old, local copy of owltools on QM’s HPC cluster. If it breaks/deprecated, the pipeline will need major changes. Might need to use semsimian

24 of 25

Acknowledgements

Damian Smedley
Pilar Cacheiro
Tomasz Konopka
International Mouse Phenotyping Consortium