Transforming mental disorders research with big data Ole
Transforming mental disorders research with big data Ole A. Andreassen MD, Ph. D Co. E NORMENT, Division of Mental Health and Addiction, Oslo University Hospital & Institute of Clinical Medicine, University of Oslo
Background big data mental illness • Most common diseases are complex • Multiple genes and environmental factors each with small effects «polygenic architecture» • Reduced genotyping costs • Chipping (genotyping) your DNA is soon cheaper than parking at your hospital • Brain imaging – new technology • No big effects • Small effects - The New Normal Paulus MP, Thompson WK. JAMA Psychiatry 2019 • Real-world data capture – descriptive diagnoses
Schizophrenia (n=240 000) 145 genetic loci (250) Polygenic architecture PGC Schizophrenia Work Group. Nature 2014 Pardinas et al. Nat Genet 2018
Bipolar disorders (n=220 000) 30 genetic loci (63) Polygenic architecture PGC Bipolar Disorder Work Group. Nat Genet 2019
Alzheimer’s disease (n=455, 258) 29 gene variants Jansen et al Nat Genet 2019
GWAS – power estimates heritability – polygenicity - discoverability Holland et al in review
Mathematical models of genetic architecture Bipolar disorder overlap Schizophrenia Frei et al. Nat Comm 2019
Extensive genetic overlap Frei, Smeland et al in review
Brain MRI – large scale neuroimaging • Structural MRI, Free. Surfer, n=65, 000 • Brain MRI from 6503 individuals • 2000 bipolar disorder • 3000 schizophrenia • 25000 healthy controls http: //enigma. ini. usc. edu/ongoing/enigma-bipolar-working-group/ NORMENT: Westlye - Agartz
Cortical thinning in bipolar disorder compared to healthy controls 2, 260 BD 3, 819 controls No differences between BD 1 and BD 2 Hibar et al. , 2018. Mol Psychiatry
Cerebellar volume and cerebellocerebral structural covariance in schizophrenia Reduced total cerebellar volume functional connectivity with associative regions of the cerebral cortex Moberget et al. , Mol Psychiatry 2018
Common brain disorders are associated with heritable patterns of apparent aging of the brain n= 45, 615 Kaufmann et al. , Nat Neurosci 2019
Prediction Example Alzheimer’s Disease • Identify individuals at risk late-onset Alzheimer’s disease (AD) – too late if memory loss • GWAS have identified several variants associated with AD beyond APOE e 4 • When, not if • not case-control but age of onset • polygenic hazard score
Survival analyses in AD ADGC (17, 008 AD cases and 37, 154 controls) – 28 SNPs IGAP, Cox forward regression model Desikan, Fan et al. PLo. S Med 2017
Predicted AD annualized incidence PHS modifies AD annualized incidence rates Desikan, Fan et al. PLo. S Med 2017
NORMENT TSD
NORMENT: TSD usage • p 33 (TOP) • 135 users in p 33 • Disk on durable: 102 TB (used), 137 TB (quota) • Disk on cluster: 230 TB (used), 275 TB (quota) • 860 K CPU-hours used since October 1 st, 2019 • 4 M CPU-hours quota until October 1 st, 2020 • 192 cores dedicated NORMENT (high-performance computing) • p 697 (Mo. Ba) p 830 (Bup. Gen) p 8. . (Dem. Gene) • GWAS time: previous in-house cluster: 10 hours, TSD (colossus 3): 1 hour • Custom-build services (database, Github, Min. Dag, monitoring)
Mobile data collection from participants
Conclusion • Mental disorder research progress based on big data • Growing imaging & genetics sample sizes • Develop precision medicine approaches
Acknowledgements • Study participants • PGC, ENIGMA, de. CODE, UCSD, Image. Mend • NORMENT team:
- Slides: 26