1. Consideration for Planning a Candidate Gene Association Study With TagSNPs
Shehnaz K. Hussain, PhD, ScM
Epidemiology 243: Molecular Epidemiology

2. Objectives Molecular genetics primer
Databases and tools to conduct in silico analyses for tagSNP selection/prioritization
Factors influencing statistical power

3. Central dogma
A T C G
DNA
mRNA
Protein

4. What are SNPs?
More than 99% of all nucleotides are the same in all humans
1% of nucleotides are polymorphic
SNPs>> insertions-deletions
Bi-nucleotide – T (80%) A (20%)
Where do SNPs occur?
Exons
Introns
Flanking regions

5. What are haplotypes?
A haplotype is the pattern of nucleotides on a single chromosome
Two “copies” of each chromosome
The haplotype inference problem
T T C G T A
? T ? G ? A
TA TT CG GG TA AA
? T ? G ? A
A T G G A A

6. What is linkage disequilibrium?
Linkage disequilibrium (LD) describes the non-random association of nucleotides on the same chromosome in a population
One nucleotide at one position (locus) predicts the occurrence of another nucleotide at another locus
No LD LD
Another closely related concept is Linkage Disequilibrium
The technical definition for LD is as follows: blah
It is a population measure, so it is not something that is unique to an individual
Describe figures:
Here is an example where we have no LD
We have 4 chromosomes indicated by these blue lines
Lets assume we have two SNPs, one here and one here
The variant, or minor allele of the SNP is indicated by either a purple dot at position 1 or a red dot at position 2
In this example, we see four potential scenarios, which occur at equal frequencies, which indicated that we have no LD
In this next example, we have high LD, because when the variant allele of position 1 is present, so is the variant allele of position 2

7. What are markers? Disease Phenotype
Test for association between phenotype and marker loci
Test for genetic association between the phenotype and the DSL LD
Candidate gene
Marker loci (SNPs)
Disease Susceptibility Locus

8. Disease Susceptibility Locus
What are tagSNPs?
TagSNPs are a subset of all SNPs in a gene that mark groups of SNPs in LD
Avoids redundant genotyping LD LD
Marker loci (SNPs)
Disease Susceptibility Locus

9. The joint effect of tagSNPs in cytokine genes and cigarette smoking in cervical cancer risk

10. T-cell proliferation IL - 2 gene IFN γ Activated T cell Proliferation
of TH1
cells
receptor
Proliferation
of TH1 -
cells IL IL - - 2 2 IL IL - - 2 2
gene
gene IL - 2
receptor
IFN γ
gene
Activated T
Activated T - -
cell
cell

11. Background Cigarette smoking ↑ 1.5- to 3-fold cancer risk
Cigarette smoking ↓ levels of IL-2 and IFNγ (cervical and circulating)
↓ levels of IL-2 and IFNγ
HPV persistence in the cervix
Cervical neoplasia
Decreased survival from invasive cervical cancer

12. Model Cigarette smoking HPV-associated squamous cell cervical cancer
SNPs in IL-2, IL-2R, and IFNG

13. Methods Study design Subjects Data collection
Population-based case-only study
Subjects
308 Caucasian squamous cell cervical cancer cases diagnosed
Residing in 3 western Washington counties
Data collection
Structured in–person interviews
DNA isolated from buffy coats

14. Objectives Molecular genetics primer
Databases and tools to conduct in silico analyses for tagSNP selection/prioritization
Factors influencing statistical power

15. Multi-stage tagSNP design
Select reference panel
Re-sequence panel, identify SNPs (many markers, few subjects)
Choose tagSNPs
Genotype tagSNPs in main study (few markers, many subjects)

16. 1. Select reference panel
Definition
A sample of your study population
Most representative
Samples from the Coriell Repository
Ability to integrate your data with other resources
= Candidate gene SNPs
= HapMap SNPs

17. 2. Re-sequence reference panel
Amplify and Sequence DNA
Gene
PolyPhred
Phred
Phrap
(Nickerson, 1997)
(Ewing, 1998)

18. Alternatives to re-sequencing
Program for Genomic Applications (PGA)
SeattleSNPs – inflammation
NIEHS SNPs – environmental response
Innate Immunity
International HapMap Project
5 million SNPs in four ethnically distinct populations

19. 3. Choose tagSNPs (LD) Option LDSelect Tagger r2 threshold (0.80) Yes
(Carlson, 2002)
Tagger
(de Bakker, 2005)
r2 threshold (0.80)
Yes
SNP exclusions/inclusions No
SNP design score

20. LDSelect output for IL-2 SeattleSNPs, r2≥0.80, MAF ≥0.05, Caucasians
Bin
Total Number of Sites
TagSNPs 1 2
rs rs
rs rs 3
rs rs 4 rs

21. Genomic context Exons (cSNPs) Upstream flanking region
SIFT (Ng, 2002)
PolyPhen (Ramensky, 2002)
Upstream flanking region
Intron-exon junctions

22. Sequence conservation
UCSC Genome Browser, PhasCons (Siepel, 2005)
Score
Repeat region
Unique region

23. Objectives Molecular genetics primer
Databases and tools to conduct in silico analyses for tagSNP selection/prioritization
Factors influencing statistical power

24. Minor allele frequency and genetic model
300 cases, 300 controls, alpha=0.05

25. Sample size requirementLD
SNPs genotyped
SNPs not genotyped r2
Sample size requirement
S1 and S2 -
600 S1 S2
1.00
0.85
706 S1 S2
N/r2 (Pritchard, 2001)

26. Genotype error Generally non-differential
Reduces your power
Every 1% increase in genotyping error rates requires sample size increased by 2-8% (Zou et al, 2004, Genetic Epidemiology)
Depends on error model

27. Power calculators Quanto htPowercc
G, E, G X E, G X G
Case-control, case-sibling, case-parent, and case-only designs
Quantitative or binary outcome
htPowercc r2
Power for Association With Error (PAWE)
Genotyping errors

28. TagSNP summary
Efficient yet comprehensive coverage of the genetic variation in our candidate genes
Reduce costs
Preference should be given to putatively functional variants:
Literature, gene context, sequence conservation
Influences of statistical power:
MAF, genetic model, LD, and genotyping error

29. Programs for Genomic Applications
SeattleSNPs,
NIEHS,
Innate Immunity,
International HapMap,
Coriell cell repository,
cSNP predictive analysis:
SIFT,
PolyPhen,
Vista,
The following programs can be found at the Rockefeller site,
Tagger
LDSelect
PAWE
Quanto