HapMap data for genome-wide disease association studies

1
HapMap data for genome-wide disease association
studies cases from SNP Research Center, RIKEN
Toshihiro Tanaka SNP Research Center, RIKEN
2
Millennium SNP projects in Japan (April, 2000
March, 2005)
I. Infrastructure a) collection of gene-based
SNPs 190,000 variations identified in two years
b) high-throughput genotyping system low cost,
semi-automated using Invader assay II.
Application Identification of genes with
medical importance Disease associated
genes Genes defining drug sensitivity
3
Two-step genotyping strategy for genome-wide
approach
1. genotype small number of samples (100 200)
for a large set of SNPs (100,000
250,000) 2. set p-value threshold to take
further steps (0.01) 3. loci that passed the
threshold will be further examined
by expanding the sample scale
And, also candidate gene approach
4
SNP Research Center, RIKEN
Laboratory for Cardiovascular Diseases lymphoto
xin-a (Nature Genetics, 2002) galectin-2
(Nature, 2004) Laboratory for Rheumatic
Diseases PADI4 (Nature
Genetics, 2002) SLC22A4 (Nature
Genetics, 2002) FCRL3 (Nature
Genetics, 2005) Laboratory for Bone Joint
Diseases asporin (Nature
Genetics, 2005) CILP (Nature
Genetics, 2005) CALM1 (Hum Mol
Genet, 2005) Laboratory for Diabetic
Nephropathy SLC12A3 (Diabetes,
2003) WNT5B (Am J Hum Genet,
2004) Laboratory for Allergic Diseases CLCA1
(Genes and Immunity, 2004) DAP3
(J Hum Genet, 2004) IFNA
(Hum Genet, 2004) ADAM33
(Clin Exp Allergy, 2004)
5
Purpose
To know the practical usefulness of HapMap data
for disease association studies
Question Could we have identified
disease-associated loci/SNPs if we had used SNP
data and software from HapMap HP to select SNPs
to be genotyped in the first stage screening?
6
Question, in other words.
Imagine a researcher wishing to identify certain
disease associated loci by GWA study, without
knowing any previous association reports.
He/she decided to select SNPs to be genotyped
by using HapMap data and Haploview software.
He/she examined 500 patients and 500 controls.
He/she set the threshold p-value, 0.01.
Could he/she detect loci that were previously
reported by us? (even when the associated SNPs
were hidden from HapMap data)
7
Study protocol
Obtain genotyping data around the
disease-associated loci from HapMap home
page Select tag SNPs using Haploview software
(block-by-block basis, and
Tagger) All the disease-associated SNPs were
in the database. treated as untyped
(hidden SNPs). Default settings were used for
Haploview in most conditions. Genotype selected
tag SNPs and perform association analysis for
500 case and 500 control samples
8
LGALS2 locus (candidate gene approach)
association result p4.5x10-6 OR1.23 n2,000
tagged SNPs block-by-block basis 8,9,10 Tagger
(r2gt0.8) 9,10 Tagger (r21) 9,10,11,12
9
Association analyses (comparison of allele
frequency)
SNP8
SNP10
SNP11
SNP12
SNP9
case control
M 738 676
m 262 324
case control
M 601 547
m 399 453
case control
M 696 635
m 304 365
case control
M 715 661
m 285 339
case control
M 619 551
m 381 449
P 0.0023 OR 1.35 r2 0.832 D 0.956
P 0.015 OR 1.25 r2 0.587 D 0.978
P 0.0038 OR 1.32 r2 0.867 D 0.978
P 0.0092 OR 1.29 r2 0.863 D 0.931
P 0.0020 OR 1.32 r2 0.616 D 0.973
SNP14 (disease associated SNP, MAF35.0)
case control
M 724 650
m 276 350
P 0.00036, OR 1.41
10
Haplotype association
Haplotype
8-9-10 case control chi-square p value
M-M-M 59.89 54.70 5.52 0.0187
m-m-m 25.99 32.30 9.61 0.0019
M-m-M 9.50 8.70 0.39 0.5340
M-m-m 4.41 4.20 0.05 0.8255

11
LTA locus (HLA region, genome-wide approach)
association result p3.3x10-6 n1,000
association result p1.3x10-4 n1,000
r20.866 D'1
12
Association analysis
SNP18 disease associated SNP MAF34.1
SNP9 (MAF32.5)
case control
M 581 650
m 419 350
case control
M 607 675
m 393 325
P 0.0015, OR 1.35
P 0.00033, OR 1.40
r2 0.90, D' 0.99
13
Newly identified locus for one common
disease (candidate gene approach)
association result p3.3x10-7 n3,000
100kb
no haplotype block no related SNP
14
Sample scale and cut-off p value
p value
Minor Allele Frequency 0.35
OR 1.41
OR 1.35
number of samples
15
Summary
All disease-associated SNPs were in the
database. in part, good luck, in part, good
quality of the database.
If they are treated as untyped (hidden SNPs),
we lose some of the
disease-associated loci,
depending on their haplotype structure.
There is a need to examine certain number of
samples and to set appropriate p-value
threshold to detect them, which,
naturally, should take cost of the study into
account.