Mouse phenotypes anddisease genomics

Caleb WebberMRC Functional Genomics Unit,

Oxford University, UK

No conflicts of interest to declare

Why are dispersed loci involved in similar phenotypes?

PHENOTYPE

Comparing Patient Phenotypes - evidence

Patient 1Phenotype APhenotype FPhenotype NPhenotype W

VSPatient 2Phenotype CPhenotype FPhenotype OPhenotype Z

Patient 1Phenotype A

?Phenotype F

?Phenotype NPhenotype W

?

Patient 2

?Phenotype CPhenotype FPhenotype O

??

Phenotype Z

Comparing Patient Phenotypes - language

Patient 2

Phenotype CPhenotype FPhenotype O

Phenotype Z

Patient 1Phenotype A

Phenotype F

Phenotype NPhenotype W

Phenotype Z Phenotype W

Human Phenotype Ontology

“Pathways” and convergent phenotypesHPO phenotypes

~4,000 patientswith developmental

abnormalities

systematically phenotyped and

annotated using the Human Phenotype

Ontology (HPO)

Genes affected by copy number variants

in these patients genomes

Functional enrichment

tests

“Pathways” and convergent phenotypes

All patients with a specific

phenotype

Patients with the specific henotype whose variant genes form part of a particular “pathway”:

Q: Are these patients more phenotypically-

similar?

p

Similarity in patients whose variant genes contribute to the same “pathway” term

phenotypically similar patients

share term

Random

phenotypically dissimilar

patients share term

PMID: 25781962

Same pathway = patients share a more specific phenotype?

A

A1 A2

A1a A1b A2a

B

B1 B2

B1a B1b B2a

C

C1 C2

C1a C1b C2a

Same “pathway” = more specific phenotype

Same “pathway” = similar patterns of phenotypes

Pathway associated phenotype

Patients whose variant genes contribute to the same “pathway” term – All phenotypes vs narrow phenotypes

phenotypically similar patients

share term

Random

phenotypically dissimilar

patients share term

PMID: 25781962

Same pathway = same pattern of phenotypes

A

A1 A2

A1a A1b A2a

B

B1 B2

B1a B1b B2a

C

C1 C2

C1a C1b C2a

Same “pathway” = similar patterns of phenotype

Take homes

1. Patients whose variants disrupt the same “pathway” share a broad range of phenotypic similarities

Functionally-linking genes through orthogonal data sources

protein interaction from high-

throughput study

protein interaction from low-

throughput study

Similar gene annotations

Mouse models have similar phenotypes

Genetically-interact in yeast

What does functionally-similar mean?

Do I trust these experiments equally?

What is the chance of seeing this by random?

Functional-linkage networks: Integration of functional genomics resources to identify

human disease genes

A B C

D E F

a gene

Protein interactiongene co-expressionintegrated

Predicting human phenotypic associations

Increasing similarity in “functionality” between genes

Pair of genes

Pair of genesUninformativeRandom

human disease phenotypes

associated with both genes are

similar

human disease phenotypes

associated with both genes are

dissimilar

Ability to predict whether 2 genes are involved in the same disorder

The value of mouse phenotypic data (MGI) compared to Gene Ontology (GO) annotations

Increasing similarity in functional annotation (either GO or MGI) between genes

Predict they are

Predict they are not

Know nothing

Weighting functional informationThe data types are assessed and weighted according to how well they predict

shared mouse phenotypesMouse orthologue KO

phenotypes

more simila

less simila

Gene expression

r

r

PMID: 25166029

Comparison of functional data sources

Mouse orthologue KO phenotypes

more similar

less similar

Frank HontiPMID: 25166029

Comparison of functional data sources

Mouse orthologue KO phenotypes

more simil

less simila

ar

r

Frank HontiPMID: 25166029

Phenotypic-linkage networks

Genes

Phenotypic links

Exome Variant Clustering

Frank HontiPMID: 25166029

Disorder-specific networks – tuning data to specific disorders

Type 2 Diabetes relevant phenotypes

more similar diabetes-relevant

phenotypes

less similar diabetes-relevant

phenotypes

Gene property

Cyn

MP:0005379 endocrine/exocrine gland phenotypeMP:0005376 homeostasis/metabolism phenotypeMP:0005375 adipose tissue phenotype

thia Sandor

Gene overlap between exome variant sets

East Asian South Asian African

AmericanHispanic

Clustering T2D exome variants from 12,884 individuals

5 ethnic samples: South Asian, East Asian, Hispanic, African American, European

Lipid and glucose homeostasis

Take homes

1. Patients whose variants disrupt the same “pathway” share a broad range of phenotypic similarities

2. Use the mouse phenotypic data to evaluate other functional data, especially for particular phenotypes of interest

Steinberg et al, NAR,

Pred

ictio

n A

ccur

acy

(MC

C)

Study Bias in haploinsufficiency prediction

Human Data

Mouse Data

Take homes

1. Patients whose variants disrupt the same “pathway” share a broad range of phenotypic similarities

2. Use the mouse phenotypic data to evaluate other functional data, especially for particular phenotypes of interest

3. We need less studied genes phenotyped to help our estimates of variant deleteriousness

Disease Genomics Programme, FGU, Oxford University

Webber Group

Stuart Grice Avi Taylor Cynthia Sandor Steph Millin Tallulah Andrews

Viola Volpato Frank Wessely Kieran Campbell Katarina Vrcelj Julia Steinberg

Diabetes exomes in collaboration with Mark McCarthy;Nijmegen CNVs with Bert DeVries