SNP selection for candidate gene association studies

1
SNP selection for candidate gene association
studies

• Peter Kraft (pkraft_at_hsph.harvard.edu)
• Rulla Tamimi
• (nhrmt_at_channing.harvard.edu)

2
Background

• Ataxia telangiectasia (AT) is an autosomal
  recessive disorder
• characterized by neurodegeneration, cerebral
  ataxia, sensitivity to radiation
• 100-fold increased risk of developing cancer
• Family studies suggest that women heterozygous
  for the ataxia telangiectasia mutated (ATM) gene
  are estimated to have 4-5 fold increased risk of
  breast cancer

3
Background

• ATM is a tumor suppressor gene
• Functions in cell cycle arrest, apoptosis and
  repair of double strand breaks
• In vitro evidence indicates that cells from AT
  heterozygotes are intermediate in their
  sensitivity to X-Rays
• gt200 disease causing mutations, mostly truncation
  mutations
• ATM mutations observed in breast cancer patients
  are mostly missense mutations

4
Epidemiologic Studies

• Studies examining polymorphic variation in ATM
  and breast cancer have been inconclusive
• Increased risk among women with family history
  and/or early onset
• Not all studies confirm these findings
• 2 hospital-based studies reported positive
  associations with missense mutations
• Population based study provided little support
  for a role of 20 ATM missense mutations in breast
  cancer

5
ATM

• Large gene
• 66 exons, over 180kb
• Two independent groups identified variation in
  this gene
• Thorstenson et al. American Journal of Human
  Genetics, 2001
• Bonnen et al. American Journal of Human Genetics,
  2000

6
Thorstenson et al.

• SNP discovery
• Used DHPLC on 93 cell lines of diverse ethnicity
• 62 out of 66 exons, splice sites, 5 and 3
  regions
• 24kb covered
• 88 variants found
• 53 found 1 time only
• 18 2-3 times
• 17 gt3 times
• Genotyped same population for these 17 SNPs
• 10 in complete LD
• 7 total haplotypes (3 in European Caucasians)

7
Thorstenson, et al. 2001
8
Bonnen et al.

• SNP discovery
• Sequenced 500bp regions evenly distributed
  throughout the gene in 5 unrelated European
  Caucasians
• Primarily noncoding regions
• 15kb sequence covered
• 17 variants found
• Sequenced population for 14 SNPs
• 71 African Americans, 77 European White
  Americans, 73 Hispanics, 39 Asian Americans
• 22 total haplotypes (only 7 in European
  Caucasians, 5gt5)

9
ATM gene
Bonnen et al, Am J Hum Genet, 2000
10
Bonnen et al, Am J Hum Genet, 2000
11
Objective

• Determine if common variation in ATM is
  associated with breast cancer
• Because of the size of ATM and apparent small
  numbers of haplotypes
• small number of htSNPs
• A haplotype approach may be useful method to
  examine common genetic variation
• Based on Bonnen paper 5 htSNPs required to
  capture 99 variation

12
Haplotype Tagging SNPs

• Using BEST
• http//genomethods.org/best/
• Uses an exact method to identify the minimum
  number of tagging SNPs necessary to capture
  variation
• BEST identified 5 htSNPs necessary to capture all
  of the haplotypes occurring at gt1

13
If the space below a column is blank, the SNP
belongs to the minimum set of htSNPs tagging the
haplotypes if a column is labeled with an X, it
can be derived from the selected htSNPs when a
column is labeled by a number, the representation
binary representation of the labeled and the
labeling columns are identical and therefore any
solution including one will be equivalent to any
solution including the other.
14
Methods

• Conducted nested case-control study in the NHS
• 1309 breast cancer cases, 1761 controls
• Genotyped 5 htSNPs using Taqman
• Used PROC HAPLOTYPE to estimate haplotypes in
  cases and controls separately
• Excluded haplotypes at lt1
• Conducted secondary analysis among those with no
  missing genotype data(1199 cases, 1535 controls)

15
Results

• Six unique haplotypes
• Haplotypes 1,4,5 represent gt80 of haplotypes
• Global p-value-0.63

16
Proc Haplotype vs. HAPPY
17
Additional Results

• Used expectation substitution approach to examine
  haplotype interactions (HAPPY)
• No interaction between haplotypes
• Family history (p0.51)
• Menopausal status at diagnosis (p0.29)

18
Limitations

• Accuracy of the estimated haplotypes relies on
  the precision of the 5 htSNPs
• No one has resequenced ATM entirely
• Efficiency depends on the density of markers used
  to choose tagging SNPs
• Bonnen markers average density of 1 SNP per 10kb
• Does not exclude possibility of rare variants
  associated with breast cancer

19

• S49C Among carriers 50 had haplotype 2, not
  significantly different from noncarriers (38
  p0.6)
• D1853N haplotype 15 only occurred among carriers
• P1054R Among carriers 46 had haplotype 17,
  significantly more than noncarriers (9 p0.003)

Letraro et al, Am J Hum Genet, 2003