Towards reconstructing personalised causal regulatory networks using - - PowerPoint PPT Presentation

▶
towards reconstructing personalised causal regulatory
SMART_READER_LITE
LIVE PREVIEW

Towards reconstructing personalised causal regulatory networks using - - PowerPoint PPT Presentation

Aug 24, 2023 198 likes •598 views

Towards reconstructing personalised causal regulatory networks using large-scale trans -eQTL and single-cell co-expression QTL analysis Annique Claringbould Department of Genetics University Medical Center Groningen Slides Lude Franke Twelve

slide-1
SLIDE 1

Towards reconstructing personalised causal regulatory networks using large-scale trans-eQTL and single-cell co-expression QTL analysis

Annique Claringbould Department of Genetics University Medical Center Groningen Slides Lude Franke

slide-2
SLIDE 2

Twelve years of genome-wide association studies

Effect: Cause:

slide-3
SLIDE 3

Genes_unknown Pathways_unknown Cell-types_unknown >10,000 known >200 diseases Genetic risk factors Disease

Black Box

The black box challenge

slide-4
SLIDE 4

Far majority of genetic risk factors affect gene expression

genetic risk factor for type 1 diabetes

Dubois et al, Nature Genetics 2010 Westra et al, Nature Genetics 2013 Fehrmann et al, Nature Genetics 2015 Zhernakova et al, Nature Genetics 2017

B-Cell

gene X expression >

T-Cell Monocyte P = 0.9

CC CT TT

P = 0.8

CC CT TT CC CT TT

P = 10-9

Cell-type specific cis-eQTL effect:

slide-5
SLIDE 5

Effects of genetic variants on single- cell gene expression

slide-6
SLIDE 6

CD4+ T CD8+ T CD56(dim) NK CD56(bright) NK cMonocyte ncMonocyte B Plasma mDC pDC Megakaryocyte HPC 1 2 3 4 5 6 7 8 9 10 11 12 t-SNE 1 t-SNE 2

scRNA-seq analysis in 25,000 PBMCs (45 different individuals)

Monique van der Wijst et al, Nature Genetics 2018

slide-7
SLIDE 7

rs4821670 affects LGALS2 in cis:

Single-cell cis-eQTL analysis

rs9332431 affects CHTF8 in cis: Monique van der Wijst et al, Nature Genetics 2018

slide-8
SLIDE 8

Downstream effects using 31,684 samples

slide-9
SLIDE 9

Goal

Disease SNP Disease SNP Disease SNP Disease SNP Disease SNP Y D E Tissue 2

Genome-wide association studies

A B X C Tissue 1

cis-eQTL effects: trans-eQTL effects:

cis-eQTL mapping trans-eQTL mapping

Z Disease

Key driver gene

Key driver gene identification

slide-10
SLIDE 10

Get larger sample-sizes: meta-analysis in 5,311 samples

Westra et al, Nature Genetics 2013 Zhernakova et al, Nature Genetics 2017 Bonder et al, Nature Genetics 2017

Downstream trans-eQTL effects

  • Chr. 7
  • Chr. 7

Systemic lupus erythematosis risk factor: Local expression effect: Type 1 interferon response: IKZF1 (in Monocytes) MX1 IFIT1 IFI44L IFI6

2 1 3 D

  • w

n s t r e a m e f f e c t s i d e n t i fi e d 
 f

  • r

3 4 6 g e n e t i c r i s k f a c t

  • r

s 2 1 8 A i m t

  • fi

n d m

  • r

e , u s i n g m a n y m

  • r

e s a m p l e s

slide-11
SLIDE 11 networks trans-eQTLs genes genetic variants

genenetwork.nl/eqtlgen Large-scale eQTL analysis: eQTLGen

www.eqtlgen.org Large-scale eQTL analysis: 37 population based cohorts Genotype data and gene expression in blood available 31,684 samples

eQTLGen

slide-12
SLIDE 12

Large-scale eQTL analysis: eQTLGen

trans-eQTL analysis: 10,317 trait-associated SNPs studied trans-eQTL analysis results: 6,298 (31%) trans-eQTL genes 3,853 (36%) genetic risk factors Polygenic score analysis results: 2,658 (13%) eQTS genes 689 (54%) traits afgect gene expression cis-eQTL analysis results: 16,989 (88.3%) cis-eQTL genes 238,340 unlinked cis-eQTL SNPs cis-eQTL analysis: 11M SNPs studied (Window size 1Mb, MAF ≥ 1%) Polygenic risk score analysis: 1,263 traits studied

A

eQTLGen Consortium

31,684 blood samples

' 5 ' 3 19,960 genes studied

11M SNPs (MAF ≥ 1%) 10,317 trait-associated SNPs Gene A Gene B Gene C Disease SNP Polygenic risk for disease > Gene expression

trans-eQTL efgects Susan Peter Kate John

X Y Y Z Disease SNP A

cis-eQTL efgect

Võsa et al, BioRxiv 2018

slide-13
SLIDE 13

Expression levels of nearly every gene are influenced by SNPs

cis-eQTLs

Nearly every gene is showing a significant cis-eQTL effect

−5 0.0 0.2 0.4 0.6 0.8 1.0

P = 0.54 P = 0.002 P = 0.22 P = 0.02 P = 0.95 P = 4 x 10−4 P = 0.09 P = 2 x 10-7 P = 0.15 P = 2 x 10-6 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

Average blood gene expression

Low High

Proportion of genes showing cis-eQTL effect Proportion of genes pLI score

slide-14
SLIDE 14

Loss of function intolerant genes

−5 −4

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

Average blood gene expression

Low High

Proportion of genes showing cis-eQTL effect Proportion of genes

34%

66%

Gene showing no eQTL effect in blood, but showing eQTL in GTEx Genes showing no eQTL effect in eQTLGen nor in GTEx

Enriched pathway Carcinoma RNA processing RNA splicing P-Value 5 x 10-11 2 x 10-9 2 x 10-7

slide-15
SLIDE 15

96.2% of lead eSNPs map within 100kb of cis-gene

slide-16
SLIDE 16

Limited evidence blood cis-eQTLs pinpoint disease genes

Analysis of cis-eQTLs using SMR for 16 well-powered traits Prioritized SMR genes do not overlap more often than expected with genes, prioritised using pathway enrichment method DEPICT

slide-17
SLIDE 17

37% of genetic risk factors for disease affect expression in trans

trans-eQTLs

37% of 10,000 risk factors affect gene expression levels in trans

Average blood gene expression

Low High

showing trans-eQTL effect

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

0.0 0.2 0.4 0.6 0.8 1.0

pLI score

P = 0.54 P = 9 x 10-4 P = 0.78 P = 0.95 P = 0.13 P = 0.10 P = 3 x 10-6 P = 10−5 P = 0.005 P = 6 x 10-7

−4
slide-18
SLIDE 18

Biological mechanism of trans-eQTLs

Biological mechanism known for trans-eQTLs Biological mechanism unknown for trans-eQTLs

Susceptibility locus trans-eQTL gene X Disease SNP Susceptibility locus

Transcription factor binding

trans-eQTL gene X Disease SNP B

Co-expression

A Susceptibility locus trans-eQTL gene X Disease SNP B

Protein-protein interaction

A Disease SNP trans-eQTL gene X

Close physical proximity

Fold enrichment = 1.98x, P = 4 x 10-83 (Co-expression based on 31,684 eQTLGen samples) Fold enrichment = 1.19x, P = 0.05 (Protein interactions based on InWeb) Fold enrichment = 2.2x, P = 2 x 10-61 (RegulatoryCircuits, Marbach et al, Nature Biotechnology 2016) Fold enrichment = 0.99x, P = 0.30 (Hi-C interactions, Rao et al, Cell 2014) TF A Susceptibility locus trans-eQTL gene Y X Disease SNP

Indirect transcription factor binding

Fold enrichment = 3.2x, P < 10-300 (RegulatoryCircuits, Marbach et al, Nature Biotechnology 2016, co-expression based on 31,684 eQTLGen samples) TF A co-expression Susceptibility locus trans-eQTL gene Disease SNP

Expression levels of local gene mediate trans-eQTL efgect

Fold enrichment = 5.3x, P = 10-67 (Tested in 3,831 BIOS samples) B A

3% 1 % 83%

X

4%

slide-19
SLIDE 19

Using blood trans-eQTLs to gain insight into brain genes

rs17087335 REST cis trans

CAD SNP affects REST transcription factor: Trans-genes specific for brain:

slide-20
SLIDE 20

Trans-eQTL effects in cancer

trans-eQTL efgects trans-eQTL efgect trans-eQTL efgect cis-eQTL efgects

ATF7IP rs116766442 rs4487645 Multiple myeloma risk factor Multiple myeloma risk factor O-glycan biosynthesis GOLM1

trans-eQTL efgect trans-eQTL efgect

rs7745098 rs114865495 Hodgkin’s lymphoma risk factor Hodgkin’s lymphoma risk factor Cell cycle RTKN2 rs2900333 Testicular germ cell tumor risk factor Chromatin organization DNA repair Gametocyte specifjc factor 1 Highly expres- sed in testis Male meiosis Male meiosis / Highly expressed in testis GTSF1 FAM50B DDX43

trans-eQTL efgects missense variant

ATM rs1801516 Melanoma risk factor DNA Damage / Telomere Stress Induced Senescence HIST1H2AC H1F0 HIST2H2BF HIST1H2BC HIST1H2BE HIST2H2BE HIST1H1PS1 HIST1H4E HIST1H2BD HIST1H1C HIST1H4H HIST1H3D

slide-21
SLIDE 21

Converging effects in systemic lupus erythematosis

rs7097397 rs2111485 rs1990760 rs4917014 rs877819 rs2663052 rs17849501 rs597808 rs9888739 rs11574637 rs35472514 rs34572943 rs1143679 rs10774625 rs1913517 ISG15 IFI6 IFI44L IFI44 RSAD2 HERC5 IFIT1 OAS3 OAS2 OASL EPSTI1 MX1 16p11.2 10q11.23 12q24.12 7p12.2 1q25.3 2q24.2

> Polygenic SLE Risk > > Expression of interferon genes

EIF2AK2 OAS1 DDX58 OAS3 OAS2 IFIT3 IFIT2 IFI6 HELZ2 XAF1 EPSTI1 RSAD2 CMPK2 OASL IFI44L IFI44 PARP9 HERC5 MX1 PARP14 IFIT1

Interferon genes

slide-22
SLIDE 22

What is a polygenic risk score?

403 variants associated to type 2 diabetes

Alice Bob Carl

11 risk alleles 189 risk alleles 362 risk alleles

Polygenic scores calculated for 1,263 diseases and traits

slide-23
SLIDE 23

Genetic risk scores on several metabolites PGS correlations: PHDGH and PSAT1: enzymes in formation of serine, acetylglycine, glycine and creatine

Glycose 3-Phosphoglycerate Pyruvate 3-Phosphoserine L-serine Phospholipids 3-Phosphohydroxypyruvate PSAT PSPH SHMT Glycine Creatine N-acteylglycine Derivative of glycine Glycine is upstream on biosynthetic pathway 3-PGDH ASNS ANKRD36BP2 CHRM3-AS2 ALKBH7 SLC7A1 FBXO9 RP11-439E19.8 ANKHD1 AARS PSAT1 PHGDH L-serine Glycine Creatine N-acteylgycine

slide-24
SLIDE 24

HDL Cholesterol: genetic risk score correlation

ABCA1 ABCG1 Apo A-1 LDLR Mature HDL Nascent HDL SR-BI LDL / VLDL CETP Cholesterol SREBP2 (SREBF2)

Foam cell Liver

slide-25
SLIDE 25 CD8+ T-Cells

Celiac disease

CD56 (dim) NK Cells CD56 (bright) NK Cells Plasma cells mDCs pDC Megakaryocytes Expression enrichment:

Cell-types with signifjcant correlation to ePRS results (BLUEPRINT consortium): Single peripheral blood mononuclear cells with signifjcant correlation to ePRS results:

CD14+, CD16- classical monocyte Macrophage Infmammatory macrophage Monocyte Osteoclast Efgector memory CD8+ αβ T cell Efgector memory CD4+ αβ T cell CD3+, CD4-, CD8+ double positive thymocyte Central memory CD4+ αβ T cell CD4+ αβ thymocyte CD8+ αβ thymocyte CD4+ αβ T cell 9.8 9.8 9.6 9.2 8.3
  • 7.5
  • 7.6
  • 7.7
  • 8.1
  • 8.1
  • 8.4
  • 10.6
Classical monocytes B cells Non-classical monocytes CD4+ T-Cells
  • 4
+4

Celiac disease polygenic risk score correlation on expression

slide-26
SLIDE 26

eQTLGen summary

Võsa et al, BioRxiv 2018

Blood cis-eQTLs have different genetic architecture as compared to disease-associated SNPs Trans-eQTLs are more informative to gain insight into downstream consequences eQTS effects are less abundant, but can help identify core genes

Cell type specifjcity > biological relevance for disease > Low High Intermediate High Low Inter- mediate cis-eQTLs trans-eQTLs eQTS 16,989 genes 6,298 genes 2,568 genes

slide-27
SLIDE 27 networks trans-eQTLs genes genetic variants

genenetwork.nl/eqtlgen

Võsa et al, BioRxiv 2018

Preprint available at BioRxiv 
 (doi.org/10.1101/447367)





All summary statistics (including 
 non-significant results) available for 
 downloading at www.eqtlgen.org


eQTLGen summary

slide-28
SLIDE 28

Identification

  • f regulatory

networks

slide-29
SLIDE 29

SREBF2 ChIP-seq binding

cis-eQTL effect on FADS2 mediated by SREBF2

Zhernakova et al, Nature Genetics 2017

FADS2 expression SREBF2 expression rs968567: A/A A/G G/G r2=0.51 r2=0.28 r2=0.10

slide-30
SLIDE 30

cis-eQTL effect on FADS2 mediated by SREBF2

SREBF2 rs968567 FADS2

Gene Gene SNP

slide-31
SLIDE 31

Gene Gene SNP Gene Gene SNP Gene Gene SNP Gene Gene SNP Gene Gene Gene Gene Gene Gene Gene Gene

Regulatory network reconstruction

slide-32
SLIDE 32

Disease SNP in gene Disease SNP in gene cis-eQTL Disease SNP cis-eQTL Disease SNP Disease SNP trans-eQTL Disease SNP in gene Disease SNP in gene cis-eQTL Disease SNP cis-eQTL Disease SNP Disease SNP trans-eQTL Disease SNP in gene Disease SNP in gene cis-eQTL Disease SNP cis-eQTL Disease SNP Disease SNP trans-eQTL

Key driver gene

From GWAS to key driver genes

Pers et al, Nature Communications 2015

slide-33
SLIDE 33

Inferring relationships using single-cell eQTLs

CC CT TT

RPS26 RPL32

r = 0.07

Expression data imputed with MAGIC§

van der Wijst et al, Nature Genetics 2018

slide-34
SLIDE 34

Inferring relationships using single-cell eQTLs

CC CT TT

RPS26

Expression data imputed with MAGIC§

RPL32

rs7297175 T/T genotype: 9 individuals rs7297175 C/C genotype: 12 individuals

  • 0.42
  • 0.78
  • 0.43
  • 0.23

0.12

  • 0.61

0.02 0.04

  • 0.63
  • 0.14
  • 0.77
  • 0.82

0.94 0.69 0.78 0.92 0.92 0.88 0.89 0.91 0.79

Correlation between RPS26 and RPL32 rs7297175 r = 0.84

  • 1
  • 0.5

0.5 1 C/C C/T T/T

slide-35
SLIDE 35

Replication of effect in whole-blood bulk RNA-seq data RPS26 RPL32 RPS26 RPL32

CC CT TT CC CT TT

scRNA-seq bulk RNA-seq (4,200 samples)

r = 0.00 r = 0.03 r = 0.26

interaction effect p = 1.3 x 10-8

van der Wijst et al, Nature Genetics 2018

slide-36
SLIDE 36

Drug response depends on your regulatory network

Disease SNP Disease SNP Disease SNP cis-eQTL cis-eQTL cis-eQTL

Key driver gene

Disease SNP Disease SNP Disease SNP cis-eQTL cis-eQTL cis-eQTL

Drug No disease symptoms Drug cures symptoms Key driver gene

Disease SNP Disease SNP Disease SNP cis-eQTL cis-eQTL cis-eQTL

Key driver gene

Disease SNP Disease SNP Disease SNP cis-eQTL cis-eQTL cis-eQTL

Key driver gene Disease symptoms Drug does not cure symptoms Drug

John Kate

slide-37
SLIDE 37 networks trans-eQTLs genes genetic variants

genenetwork.nl/eqtlgen

Consortium aim: 
 Perform large-scale single-cell eQTL meta- analysis in peripheral blood mononuclear cells Build personalised co-expression networks and identify co-expression QTLs Apply these models to individual genomes, predict cell-type specific gene expression levels

Single-cell eQTLGen Consortium

slide-38
SLIDE 38

Large-scale datasets necessary to better understand biology behind GWAS associations Downstream effects of genetic variants more informative than local effects Combined genetic risk score associations can identify key driver genes Single-cell data identifies the cell types at play Personalised co-expression networks can be built using these ingredients, and will ultimately aid pharmacological decision-making

Conclusion

slide-39
SLIDE 39

Acknowledgements >

Funding >

UMC Groningen BBMRI-NL BIOS Consortium eQTLGen Consortium

Juha Karjalainen Patrick Deelen Marc Jan Bonder Annique Claringbould Sipko van Dam Jackie Dekens Monique van der Wijst Morris Swertz 
 Peter-Bram ’t Hoen Tonu Esko Freerk van Dijk Niek de Klein Harm-Jan Westra Urmo Võsa Dylan de Vries Harm Brugge Sasha Zhernakova Jingyuan Fu 
 Bas Heijmans Vinod Kumar Sebo Withoff Yang Li Serena Sanna Dasha Zhernakova Raúl Aguirre Cisca Wijmenga Lude Franke