U L R I K E P E T E R S , F R E D H U TC H I N S O N C A N C E R R E S E A RC H C E N T E R , U N I V E R S I T Y O F WA S H I N G T O N
Fine-mapping of obesity GWAS loci using the Metabochip in PAGE (Population Architecture
using Genetics and Epidemiology)
Design of Metabochip for anthropometric related traits
Anthropometric related MetaboChip content Replication
• 13k SNPs for BMI, WHR, WC, height, % fat mass
Fine-mapping• 41 regions, 26k SNPs
Current Study Population in PAGE and Collaborative Studies
Study nARIC 3.300MEC 3,900-5,300WHI imputed 6,300WHI genotyped 5,300GenNet 500HyperGEN 1,200Total 20,500-22,000
SUBMITTED TO PLOS GENETICS 2012
16q12.2/FTO Strongest GWAS finding for obesity-related traits
16q12.2/FTO Association with BMI
r2 based on AA r2 based on EA
1,529 SNPs across 650kb
Bioinformatic Characterization by Praveen Sethupathy, UNC
Candidate intronic regulatory elements:• rs11642015, rs17817497, rs3751812, rs17817964,
rs62033408, and rs1421085 Highly sequence-conserved elements among vertebrates:
rs3751812 and rs1421085 Predicted to have allele-specific binding affinities for
different transcription factors:• rs11642015 ->Paired box protein 5 (PAX5) • rs1421085 ->Cut-like homeobox 1 (CUX1), previously
implicated in the transcriptional regulation of FTO (Stratigopoulos, J Biol Chem 2011)
Definition of Significance Levels Different alpha-levels for different aims:A. Fine-mapping regions:
1. Fine-mapping of GWAS index SNPs• Adjust only for SNPs that are correlated with GWAS index SNP
at r2>0.2, >0.5, >0.8 in population that identified GWAS index SNP (mostly EA or Asian)
• Accounting for correlation among SNPs, e.g. by permutation or estimate # of bins
2. Search for second independent signals• Adjust for all other SNPs in the fine-mapping region (excluding
those included in #1) while accounting for correlation B. Replication/generalizationC. Pleiotropy– or analysis across the Metabochip
FTO region with correlation in EAIn total 88 SNPs are correlated at r2>0.2 with 9 GWAS index SNPs in EA (all dotes that are red, yellow, green or light blue)
GWAS hit 1
GWAS hit 2
GWAS hit 3
Example FTO Region
SNP CAF% change in BMI per coding allele Nominal p
Adjusted P
Beta estimate 95%CI Fine-mapping of GWAS index SNP (# of independent tests = 30)rs62048402 0.12 1.13 (0.51,1.74) 2.4E-04 7.2E-03rs11642015 0.11 1.09 (0.47,1.7) 4.9E-04 0.01rs56094641 0.12 1.12 (0.5,1.73) 2.8E-04 8.4E-03rs55872725 0.11 1.09 (0.47,1.7) 5.3E-04 0.02rs1421085 0.12 1.11 (0.49,1.72) 3.0E-04 9.1E-03
Search for second independent signals (# of independent tests = 1,109)rs59109276 4.3E-03 1.00rs11642841 4.8E-03 1.00rs13330831 5.5E-03 1.00
Based on ~21,000 subjects (ARIC, HyperGEN, GenNet, MEC, WHI)
Summary for primary signalsregion # SNPs SNP MAF Effect P.value adjusted P r2 in AA r2 in EA
1 1p31.1 Top rs2613504 0.19 0.007 2.65E-03 0.70 265 GWAS rs2568958 0.46 0.001 0.74 0.11 0.32
rs2815752 0.46 0.001 0.7 0.11 0.322 1p31.1 Top rs7553158 0.23 -0.007 4.56E-04 0.03
55 GWAS rs1514175 0.34 0.002 0.32 0.31 0.933 1q25.2 86 Top/GWAS rs543874 0.25 -0.012 1.01E-08 8.70E-07 4 2p25.3 Top rs111593420 0.10 0.015 5.80E-07 1.05E-04
181 GWAS rs6548238 0.12 -0.013 3.00E-06 0.64 0.94rs7561317 0.24 -0.005 0.02 0.28 0.93rs2867125 0.12 -0.012 1.13E-05 0.70 0.97
5 3p12.1 Top rs1375564 0.25 -0.007 3.05E-04 0.04 127 GWAS rs13078807 0.06 -0.001 0.8 0.00 0.30
6 3q27.2 Top/GWAS rs7647305 0.41 -0.007 1.76E-04 0.0 62 GWAS rs9816226 0.21 0.007 1.00E-03 0.36 0.85
7 4p12 Top rs348495 0.35 -0.012 2.49E-07 1.22E-05 *49 GWAS rs10938397 0.25 -0.008 1.07E-04 0.60
8 5q13.3 Top rs767676 0.19 0.012 1.03E-04 2.61E-02 253 GWAS rs2112347 0.50 0.000 0.99 0.05 0.27
9 6p12.3 Top rs2744475 0.33 -0.007 1.03E-04 1.07E-02 104 GWAS rs987237 0.11 -0.005 0.08 0.24 0.50
10 9p21.1 Top rs17770336 0.18 0.008 7.12E-04 7.48E-02 105 GWAS rs10968576 0.17 -0.007 3.10E-03 0.89 0.99
11 11p15.4 Top rs10128597 0.18 -0.014 3.41E-05 7.19E-03 211 GWAS rs4929949 0.40 -0.001 0.76 0.02 0.33
region # SNPs SNP MAF Effect P.value adjusted P r2 in AA r2 in EA12 11p14.1 top rs1519480 0.25 -0.012 9.95E-09 1.18E-06
119 GWAS rs925946 0.26 -0.002 0.35 0.12 0.97top rs35070613 0.02 0.032 2.47E-07 2.72E-05 (r2<0.1 with rs925946)
110 GWAS rs10767664 0.07 0.017 3.99E-03 * *rs6265 0.05 -0.021 5.80E-07 0.33 0.38
13 11p11.2 Top rs6485802 0.18 -0.010 8.93E-05 2.22E-02 249 GWAS rs10838738 0.10 0.003 0.28 0.24 0.14
rs3817334 0.26 -0.001 0.52 0.04 0.0714 12q13.12 Top rs10875982 0.38 -0.004 0.04 2.52E+00
63 GWAS rs7138803 0.17 0.002 0.49 0.34 0.7915 14q12 Top rs28401580 0.33 -0.003 0.09 1.80E+00
20 GWAS rs11847697 0.33 0.002 0.3 0.51 0.9216 15q23 Top rs8025163 0.02 -0.018 9.91E-04 1.80E-01
182 GWAS rs2241423 0.37 -0.003 0.15 0.02 0.2817 16p12.3 Top rs4782282 0.22 -0.009 3.31E-05 5.27E-03
159 GWAS rs12444979 0.09 -0.003 0.31 0.01 0.4718 16p11.2 Top rs115616784 0.12 -0.008 2.74E-03 4.52E-01
165 GWAS rs7498665 0.27 -0.003 0.16 0.04 0.36rs7359397 0.09 0.000 0.97 0.01 0.36
20 18q21.32 Top rs12967135 0.27 0.009 7.32E-06 1.41E-03 *192 GWAS rs17782313 0.28 -0.008 3.87E-05 0.94
rs12970134 0.14 0.008 1.84E-03 0.13rs10871777 0.29 -0.007 1.26E-04 0.90
rs571312 0.34 0.003 0.16 0.2521 19q13.11 Top rs14810 0.15 -0.009 3.86E-04 1.00E-02
26 GWAS rs11084753 0.36 -0.003 0.27 0.22 0.62rs29941 0.18 -0.006 0.01 0.67 1.00
22 4q24 Top rs151411 0.25 -0.007 0.03 7.20E-01 24 GWAS rs13107325 0.01 0.000 0.99 0.04 0.18
11p14.1/BDNF,LIN7C,LGR4
Correlation based on EA with 2 different GWAS index SNPs
11p14.1/BDNF,LIN7C,LGR4
Correlation based on AA with one GWAS index SNP and most significant SNP in the region
r2 with GWAS hits
region region # SNPs SNP MAF Effect P.value Adjusted P r2 in EA r2 in AA1 1p31.1 862 rs114875057 0.03 -0.0274 2.31E-04 0.20 * <0.012 1p31.1 255 rs74543698 0.00 0.2736 2.85E-03 0.73 <0.01 <0.013 1q25.2 580 rs3131310 0.06 -0.0155 4.66E-05 0.03 0.03 0.014 2p25.3 727 rs2683962 0.14 -0.0096 2.16E-04 0.15 <0.01 <0.015 3p12.1 507 rs114616854 0.01 -0.033 6.26E-04 0.32 * <0.016 3q27.2 242 rs78419649 0.02 -0.0169 8.95E-03 2.17 <0.01 <0.037 4p12 240 rs116810097 0.25 -0.0087 1.89E-05 4.53E-03 * 0.118 5q13.3 647 rs80324692 0.02 0.023 9.80E-04 0.63 0.03 0.029 6p12.3 1330 rs9784814 0.32 0.0122 5.05E-06 0.01 * 0.02
10 9p21.1 189 rs16913123 0.05 -0.0162 6.63E-03 1.25 <0.01 0.0111 11p15.4 333 rs76633799 0.04 0.0172 1.81E-04 0.06 * 0.0312 11p14.1 354 rs12284158 0.24 0.0117 1.51E-08 5.36E-06 <0.01 <0.213 11p11.2 1280 rs61895765 0.01 -0.0489 1.06E-05 0.01 <0.02 <0.0114 12q13.12 201 rs114956532 0.01 -0.0329 0.01 2.30 * <0.0115 14q12 189 rs66955107 0.05 0.0129 1.26E-03 0.24 0.01 0.0216 15q23 835 rs75821692 0.03 0.0167 5.02E-04 0.42 * 0.0217 16p12.3 447 rs11644432 0.17 0.0131 1.20E-07 5.36E-05 0.05 <0.0118 16p11.2 385 rs34413922 0.03 0.0284 3.56E-04 0.14 <0.04 <0.0120 18q21.32 903 rs73445651 0.07 0.0136 2.34E-04 0.21 * <0.221 19q13.11 88 rs116981238 0.00 -0.1107 0.01 0.92 <0.02 <0.0122 4q24 238 rs72922936 0.04 -0.0178 1.28E-04 0.03 * <0.01
Summary for secondary signals
Decisions for Next Paper(s) Study populations
• Focus on AA, AA and Asian or multiethnic panel?• Data freeze
Outcome• Two separate papers for BMI and WHR/WC
Metabochip content• Focus on fine-mapping regions or entire Metabochip
content• Note, some of the most significant findings are outside of the
BMI regions, but require more complex follow up Overall timing
• We need to be fast to avoid being scooped by other groups
Study population for next papersStudy n Availability Include in next papersAfrican AmericansARIC 3.300 Yes XMEC 3,900-
5,300Yes X
WHI imputed 6,300 Yes XWHI genotyped 5,300 Yes XGenNet 500 Yes XHyperGEN 1,200 Yes XCARDIA ~500 NoCHS 800BioVU ~10,000 No HispanicWHI 5,500 Not cleanedSOL 12,000 Genotyping ongoingAsianMEC 3500 Genotyping ongoingWHI 3500 Genotyping ongoingThaiChi 10,000 Yes ?CLHNS 1,000 Yes ?
Within HDL region # 3rs6712203 is most significant SNP 1.7 x 10-10
Correlation between BMI and HDL ~ 0.2
GWAS hit in HDL region #3 is rs10195252
BMI HDL
lnBMI ~ SNP + HDL + age*sex + PC1 + PC2 HDL ~ SNP + BMI + age*sex + PC1 + PC2
Note results based on 11,792 subjects with HDL and BMI data (~55% of all with BMI in Manhattan plot)!
Extra slides
Example FTO region: Fine-mapping of GWAS index SNPs 1,529 SNP genotyped across 640kb region Correlation with 9 index SNPs in CEU (EA) 1000
Genome Project pilot:• r2>0.2 = 88 SNPs on Metabochip (r2>0.5 = 72; r2>0.8 = 59
SNPs) Permute random normal distributed phenotype and
run analysis of all 97 (88+9) SNPs 10,000 times to compute the # of independent tests =>30• Nominal p-value * number of independent test = multi-
comparison adjusted p-value (e.g. 2.4E-04*30=7.2E-03)OR• Alpha of 0.05 /# of independent test = multi-comparison
adjusted alpha level (e.g. 0.05/30 = 0.002)
1,529 SNP genotyped across 640kb• Exclude 97 SNPs included in fine-mapping of GWAS index
SNPs (1,529-97 = 1,432)• Repeat permutation for all SNPs in entire region => 1109
independent tests
Example FTO region: Search for second independent signals
There are 1,432 SNPs that are not correlated with GWAS index SNPs in EA (r2<0.2, dark blue dots)
These result in 1,109 independent tests
Exploration if most significant BMI locus is independent from HDL
lnBMI ~ SNP + HDL + age*sex + PC1 + PC2 HDL ~ SNP + BMI + age*sex + PC1 + PC2
BMI HDL
Note results based on 11,792 subjects with HDL and BMI data (~55% of all with BMI in Manhattan plot)!
Same as slide before but not mutually adjusted for HDL and BMI
lnBMI ~ SNP + age*sex + PC1 + PC2 HDL ~ SNP + age*sex + PC1 + PC2
BMI HDL
Note results based on 11,792 subjects with HDL and BMI data (~55% of all with BMI in Manhattan plot)!