evaluating mutation intolerance and natural selection in
TRANSCRIPT
Evaluating mutation intolerance and natural selection in schizophrenia
GWAS data
Antonio F. Pardiñas
The CLOZUK project
▪ Genetics of treatment-resistance in SCZ.
▪ Initiated with CLOZUK1 (N=7,000).▪ Expanded with CLOZUK2 (N=8,000).▪ Collaboration with Novartis & Leyden Delta.
▪ Anonymised blood samples genotyped.▪ Controls obtained through collaboration.
Dataset Status Samples in GWAS Reference
CLOZUK1 Cases 5,528 Hamshere et al. 2013
CLOZUK2 Cases 4,973 Pardiñas et al. BioRxiv
CardiffCOGS1 Cases 512 Rees et al. 2014
CardiffCOGS2 Cases 247 Pardiñas et al. BioRxiv
WTCCC2 Controls 4,641 WTCCC, 2007
Cardiff Controls Controls 1,078 Green et al. 2010
Generation Scotland Controls 6,480 Amador et al. 2015
T1DGC Controls 2,532 Hilner et al. 2010
POBI Controls 2,516 Leslie et al. 2015
TWINSUK Controls 2,426 Moayyeri et al. 2013
QIMR Controls 2,339 Wright et al. 2004
TEDS Controls 1,752 Haworth et al. 2013
GERAD Controls 778 Harold et al. 2009
The CLOZUK project
Meta-analysis(CLOZUK + PGC)
40,675 cases and 64,643 controls.
143 independentGWS loci(50 novel)
Gene-set analyses
Set name Source N (genes) P-value P-value (FWER- corrected)
LoF-intolerant genes ExAC 2921 4.07 x 10-16 2.02 x 10-13
* MGI gene sets curated by Pocklington et al. 2015.
Gene-set analyses
Set name Source N (genes) P-value P-value (FWER- corrected)
LoF-intolerant genes ExAC 2921 4.07 x 10-16 2.02 x 10-13
FMRP targets Darnell et al. 2011 798 1.92 x 10-8 0.00001
Abnormal behaviour MGI database* 1939 1.20 x 10-6 0.00018
Abnormal nervioussystem
electrophysiologyMGI database* 201 2.27 x 10-5 0.00303
Voltage-gated Ca++
channel complexes Müller et al. 2010 196 8.01 x 10-5 0.01144
5HT-2C receptor Becamel et al. 2002 16 2.26 x 10-4 0.02924
Abnornal long termpotentiation MGI database* 142 2.32 x 10-4 0.02982
* MGI gene sets curated by Pocklington et al. 2015.
How much signal explained by gene sets? (Gusev et al. 2016, de Leeuw et al. 2016)
Genic SNPs accounted for 64% h2
SNP
SNPs in CNS-related genes: 39% h2
SNP
SNPs in LoF-intolerant genes : 30% h2
SNP
Gene set analysisLoF-intolerant genes are more enriched than any set from a curated collection.
Summary (1)
Partitioned heritabilityLoF-intolerant genes explain 30% h2
SNP and account for enrichment in other sets.
▪ Robust enrichment of FMRP targets.
▪ Utility of LoF-intolerance (Lek et al. 2016) for highlighting genes with common risk alleles.
Risk alleles of large effect recurrently eliminated from the population (Rees 2011; Kirov 2012, 2014)
Risk alleles of small effect persist at common frequencies (ISC 2009, PGC 2014).
▪ Usually diagnosed in young adults.▪ Reduces life expectancy by 10-20 years.▪ Patients have 30% of the fecundity of the
general population (Power 2013)
▪ Effects too small to be selected against.▪ Potential (past) beneficial effects caused
these alleles to be selected for.
Why do risk alleles persist?
Positive selection in psychiatric disorders
SchizophreniaRisk alleles might be advantageous for mating (Crow 1993, Shaner 2004). Positive effects on creativity (Kyoga 2011, Power 2015).
ASDRisk alleles linked to cognition might be favoured by assortative mating (Crespi 2016) or ancient selection (Polimanti 2017).
DepressionRisk allelles linked to sun tolerance might have been benefitial for human populations outside of Africa.(Simonti 2016)
Comparing GWAS results with selection signals
▪ CLOZUK+PGC summary statistics.▪ LDSR partitioned heritability.
▪ SNP-based selection metrics:▫ iHS, CMS (~30,000 yrs BP; Grossman et al. 2013)▫ XP-EEH (~50,000 yrs BP; Sabeti et al. 2007)▫ CLR (>60,000 yrs BP; Huber et al. 2016)▫ B (Background selection; McVicker et al. 2009)
Results
Annotation Enrichment (genomic top 2%)
P-valueEnrichment
(genomic top 1%)P-value
Backgroundselection 1.801 0.001 2.341 9.9×10-4
iHS 0.973 0.946 0.980 0.974
CMS 0.053 0.001 0.037 0.006
XP-EEH 0.621 0.034 0.383 0.303
CLR 0.401 6.5×10-5 0.173 5.8×10-7
Follow-up analyses
▪ Is the result confounded by thresholding?▫ No, supported by quantitative LDSR.
▪ Is the result confounded by functionality?▫ No, supported by significant LDSR tau-C.
▪ Is there any mechanistic explanation?▪ Is it feasible for schizophrenia?
Background selection (BGS) occurs in genomic regions with low recombination and reduces genetic diversity:
Neutral alleles: Deleterious allele:
Fn FN
Nearly-neutral model (Ohta 1973)
λ = Fixation probability.Ne = Effective population size.s = Selection coefficient.
Selection is more effective in large, diverse populations.
Neutrality limit: abs(2×Ne×s) > 1
Mutations in loci with reduced effective size (i.e. by BGS) can be under the effect of genetic drift.
Reduced genetic diversity allows weakly deleterious alleles to drift to high frequencies in the population:
Neutral alleles: Deleterious alleles:
Fn FN
Liability-threshold model (Dempster 1950, Wray 2010)
k = Population prevalence.
Relates effect size (OR), fecundity and selection (s).
OR=1.05; s=-2.26×10-4
Ne = 4,500 (Gravel et al. 2011)No BGS: 2×Ne×s = -2.032With BGS: 2×Ne×s×B < -0.238
No effect for CNVs (s=-0.2)
k
Feasibility in schizophrenia (E. Santiago & A. Caballero)
SimulationsTrait reduces fecundity.Causal genetic locus.Range of effective sizes.Mutations appear with small effect sizes (OR: 1.05-1.60).Individuals are sampled from the population using actual case-control frequency.
Ne decreases SNP heritability.
Feasibility in schizophrenia (E. Santiago & A. Caballero)
SimulationsTrait does not affect fecundity.Causal genetic locus.Range of effective sizes.Mutations appear with small effect sizes (OR: 1.05-1.60).Individuals are sampled from the population using actual case-control frequency.
Ne increases SNP heritability.
Positive selectionAll metrics “depleted” in LDSR analysis. No effect in human evolutionary timescales.
Summary (2)
Background selectionEnriched in LDSR analysis. Consistent with follow-up and mechanistic hypothesis.
▪ Drift, not selection, explains the findings.
▪ Compatible with reduction in fecundity and enrichment in LoF-intolerant genes.
Psychosis team
James WaltersPeter HolmansAndrew PocklingtonValentina Escott-PriceMichael O’DonovanMichael Owen
Core team
Lucinda Hopkins
HPC team
Mark Einon
Other collaborators
Stephan RipkeNaomi WrayEnrique SantiagoArmando Caballero
CRESTAR EU-FP7
Acknowledgements
THANKS!
Any questions?
@AFPopgen
“Brain” icon by Creative Stall from The Noun Project.“Human Mind” icons by Thomas Helbig from The Noun Project.