C HANGING THE C OURSE OF H UMAN

C HANGING THE C OURSE OF H UMAN H EALTH T HROUGH B OLD P URSUITS IN S CIENCE Impact of human genetics on drug R&D Robert Plenge ASCPT March 23, 2018

Key messages § Human genetics will have greatest impact on – selecting targets – matching modality with mechanism – guiding PD biomarkers design – identifying indications for clinical trials

Key messages § Human genetics will have greatest impact on – selecting targets – matching modality with mechanism – guiding PD biomarkers design – identifying indications for clinical trials § Actionable pharmacogenetics will emerge from targets and therapies based on human genetics – only rarely emerge from unbiased searches for therapies not based on human genetics – nonetheless, it is logical to search for genetic variation in treatment response (both efficacy and toxicity)

Outline § Role of genetics in selecting new targets – allelic series model § Matching modality with mechanism – Perhaps the greatest challenge – Example: TYK 2 and autoimmunity § Population-based cohorts – An emerging tool – Phe. WAS and Mendelian randomization – Examples: IFIH 1 and PRTN 3 § Concluding thoughts

Outline § Role of genetics in selecting new targets – allelic series model § Matching modality with mechanism – Perhaps the greatest challenge – Example: TYK 2 and autoimmunity § Population-based cohorts – An emerging tool – Phe. WAS and Mendelian randomization – Examples: IFIH 1 and PRTN 3 § Concluding thoughts

What is the ideal drug target?

Human Phenotype Pick a human phenotype for drug efficacy High Low LOF GOF Plenge et al NRDD 2013 Gene function

Human Phenotype Pick a human phenotype for drug efficacy High Low LOF Gene function Nelson et al NG 2015

Human Phenotype Pick a human phenotype for drug efficacy X X High X X Identify a series of alleles with range of effect sizes in humans (but of unknown function) X Low LOF Gene function

Pick a human phenotype for drug efficacy Human Phenotype Efficacy X High X X Assess biological function of alleles to estimate “efficacy” response curve X X Low LOF Gene function

Pick a human phenotype for drug efficacy Human Phenotype Efficacy X High X X X Toxicity X Assess biological function of alleles to Assess pleiotropy estimate “efficacy” as proxy for curve ADEs response X X Low LOF Gene function

New target for drug screen! Pick a human phenotype for drug efficacy Human Phenotype Efficacy X High X X X Low Toxicity X This provides evidence for therapeutic window at the beginning of the drug discovery journey. LOF Gene function

Outline § Role of genetics in selecting new targets – allelic series model § Matching modality with mechanism – Perhaps the greatest challenge – Example: TYK 2 and autoimmunity § Population-based cohorts – An emerging tool – Phe. WAS and Mendelian randomization – Examples: IFIH 1 and PRTN 3 § Concluding thoughts

Example of allelic series model: TYK 2 § TYK 2 is an intracellular signaling molecule § Rare, complete human knockout is associated with immunodeficiency and risk of infection § Common alleles reduce TYK 2 function and protect from risk of autoimmune disease (e. g. , RA, SLE, IBD) § Same common alleles do not increase risk of infection

Allele that protects from autoimmunity (e. g. , rheumatoid arthritis) is associated with loss-offunction (Lo. F)

Same Lo. F allele has no obvious increased risk of infection

Complete TYK 2 knockout (function)

Therapeutic hypothesis: Partial inhibition of TYK 2 will protect from RA (and SLE, psoriasis) without risk of infection

But matching modality with mechanism is challenging, especially selectivity over JAKs

Tokarski et al JBC 2015

Matching modality and mechanism: allosteric modulation required for TYK 2 selectivity over JAKs Dendrou et al STM 2016 Tokarski et al JBC 2015

Outline § Role of genetics in selecting new targets – allelic series model § Matching modality with mechanism – Perhaps the greatest challenge – Example: TYK 2 and autoimmunity § Population-based cohorts – An emerging tool – Phe. WAS and Mendelian randomization – Examples: IFIH 1 and PRTN 3 § Concluding thoughts

Population cohorts as unique genetic resource UK Biobank HUNT/NTNU - Finnish Founder Population Kaiser-Permanente Geisinger Human Longevity INTERVAL Estonia Precision Medicine Initiative NIH-AMP Kaadorie Biobank Veteran’s Administration ++ + + - Saudia Arabia Genome Project Singapore Biobank Ranking (estimates 10/2015): size/quality of genetic data size/quality of phenotypic data connectivity of genetics & phenotypes setup for recalls & ph 1 trials

Excellent example published yesterday in NEJM! UK Biobank HUNT/NTNU - Finnish Founder Population Kaiser-Permanente Geisinger Human Longevity INTERVAL Estonia Precision Medicine Initiative NIH-AMP Kaadorie Biobank Veteran’s Administration ++ + + - Saudia Arabia Genome Project Singapore Biobank Ranking (estimates 10/2015): size/quality of genetic data size/quality of phenotypic data connectivity of genetics & phenotypes setup for recalls & ph 1 trials

Phenome-wide association studies (Phe. WAS) EHRs, Claims, Questionnaires, etc. Diseaseagnostic cohort Test association of selected SNP with clinical endpoints Clinical data Phe. WAS code 1 risk surrogate for efficacy risk surrogate for toxicity Phe. WAS code 2 SNP Phe. WAS code 3 Phe. WAS code 4 Phe. WAS code 5 … Phe. WAS code 500 Genetic data GWAS, exome sequencing, etc.

Phe. WAS example: IFIH 1, autoimmunity, asthma • Phe. WAS in ~800, 000 individuals from four population cohorts • Tested 25 SNPs for association with 1, 683 clinical endpoints • 10 novel associations discovered • Example: IFIH 1 LOF allele protects from autoimmunity (known) but increases risk of asthma (novel finding) • Therapeutic hypothesis: inhibiting IFIH 1 may be effective in some autoimmune diseases but may make asthma worse Diogo et al under revision Vitiligo T 1 D Psoriasis SLE Asthma UC Odds ratio

Predicted impact of therapeutic inhibition of IFIH 1 Vitiligo T 1 D Psoriasis SLE Asthma UC Beneficial effect for some autoimmune diseases, but increase risk of asthma and UC Odds ratio efficacy safety

Genetics can bridge biomarker with clinical data, establishing a causal link for drug discovery G I D vs C G I D

Mendelian randomization: nature’s clinical trial

Mendelian randomization: nature’s clinical trial

MR example: PRTN 3 and ANCA+ vasculitis • Tested 3, 622 plasma proteins in 3, 301 healthy individuals from INTERVAL population cohort • Identified 1, 927 genetic associations with 1, 478 proteins • Example: PRTN 3 Go. F allele increases PR 3 protein and increases risk of PR 3 associated vasculitis • Therapeutic hypothesis: eliminating PR 3 protein or deleting autoantibody secreting B cells may treat vasculitis Sun, Maranville et al under revision

Outline § Role of genetics in selecting new targets – allelic series model § Matching modality with mechanism – Perhaps the greatest challenge – Example: TYK 2 and autoimmunity § Population-based cohorts – An emerging tool – Phe. WAS and Mendelian randomization – Examples: IFIH 1 and PRTN 3 § Concluding thoughts

Key messages § Human genetics will have greatest impact on – selecting targets – matching modality with mechanism – guiding PD biomarkers design – identifying indications for clinical trials § Actionable pharmacogenetics will emerge from targets and therapies based on human genetics – only rarely emerge from unbiased searches for therapies not based on human genetics – nonetheless, it is logical to search for genetic variation in treatment response (both efficacy and toxicity)