1. Genome Wide Association Study (GWAS) and Personalized Medicine

2. Outline Gene discovery and personalized medicine
Family linkage-based approach
Candidate gene-based approach
Whole genome scan (Genome-wide association study)
Genome wide association study (GWAS)
Objectives and approaches
Benefits and challenges
Resources and requirements
Technologies
A case study – Genome-Wide Study of Exanta Hepatic Adverse Events

3. Human Genome Project – Hunting for disease genes
February 15 & 16, 2001 Science and Nature
Implications:
Scientific advancement
Enhanced public health
Potential social issues
Genome

4. Relationship between genes and diseases - Single Gene-Driven Diseases
Rare and familial diseases caused by mutations in a single gene (e.g., cystic fibrosis and sickle-cell anemia)
AGCT
AGGGCCTT
Genome

5. Family Linkage-Based Approach
Identify Genetic Profile Through Gene Discovery - Approaches and Technologies
Family Linkage-Based Approach
Use the linkage principle to study families in which the disease occur frequently
Identify disease-susceptibility genes in rare familial diseases
More successful for diseases caused by a single gene (e.g., Huntington’s disease)
More successful for genes strongly increasing risk
Need a well documented family tree and disease history
Successful far less likely for some heritable diseases caused by interaction of many weak genes

6. Relationship between genes and diseases - Multiple Gene-Driven Diseases
Many genes interact each to cause disease
No single gene has strong effect
Must search for multiple genes functionally involved in putative disease-associated biomedical pathways
Genome

7. Candidate Gene-Based Approach
Identify Genetic Profile Through Gene Discovery - Approaches and Technologies (cont.)
Candidate Gene-Based Approach
Process
Select genes from known disease-related pathways
Search for causative mutations in the genes
e.g., ACH/Charlotte Hobbs
Knowledge-based approach
Drawbacks:
Constrained by existing knowledge
Constrained by genes examined

8. A More Complicated Picture
Genetics loads the gun, but environment pulls the trigger
Interaction between disease genes and patients’ life style and/or environment
Genome

9. A Realistic Picture + + = Diverse responses to treatment
Same (similar) symptom
+ One-fits-all

10. Diverse response to a one-fits-all treatment
Optimal
responders
Suboptimal
responders
Non-
responders
Adverse Events

11. From One-Fits-All to Personalized Medicine
Based on patients’ genetic profile, selecting patients  treatment
Optimal
responders
Suboptimal
responders
Adverse Events
Non-
responders

12. A New Way to Determine Genetic Profile - Whole Genome Scanning
Search all possible SNPs, not mutations, in all genes;
Yah, right !
Genome

13. Genetic Profile – From Mutation to SNPs
Mutations and SNPs are both genetic variation
<1% of genetic variations are disease related, & called mutations;
Mutations considered harmful and disease related
The majority of genetic variation is not disease related (>1%),& called SNPs
SNPs comprise “harmless” genetic variation (personalized)
SNPs can be used as markers for disease genes
GWAS is searching for SNPs marking disease causing mutations

14. The Era of the Genome Wide Association Study (GWAS)
A brute force approach of examining the entire genome to identify SNPs that might be disease causing mutations
Far exceeds the scope of family linkage and candidate gene approaches
Must obtain a comprehensive picture of all possible genes involved in a disease and how they interact
Objective: Identify multiple interacting disease genes and their respective pathways, thus providing a comprehensive understanding of the etiology of disease

15. GWAS Approach Case Matched/unmatched Control Association:
Individual SNPs
Alleles
Haplotype (combination of SNPs)
Disease related:
Genes
Pathways
Loci

16. Benefits and Challenges
Challenges: the uncertainty between SNPs and the disease-causing mutation requires large sample size
2000 – 4000 sample sizes
Minimum 1000
Unfortunately, most experiments have < 500 samples
Why the enthusiasm about GWAS:
Comprehensive scan of the genome in an unbiased fashion has potential to identify totally novel disease genes or susceptibility factors
Potential to identify multiple interacting disease genes and their respective/shared pathways

17. Requirements Success factors Experimental: large sample size
Platform: accurate genotyping technology
Analysis
Comprehensive SNP maps
Rapid algorithm IT
Sophisticated IT infrastructure
Powerful computers
Expertise (NCTR)
Medical doctors (NA)
HTP genotyping platforms (NA)
Population genetics (NA)
Biostatistics (Yes)
Bioinformatics (Yes)
Statistics (Yes)

18. SNP Map
Current technology not advanced enough to encompass all SNPs; not even close
Selecting SNPs based on haplotype block
Issues related to haplotype
A SNP pattern consistent across a population
Population-dependent
Analysis method-dependent
One of the objectives of HapMap LD
Hyplotype Block
Selecting SNPs

19. Selection of SNPs for GWAS

20. High-Throughput Genotyping Technology
Several diverse technologies, but moving to array-based approaches
Array-based technologies: Illumina, Affymetrix, Perlegen and NimbleGene
Very similar to the technology used for gene expression microarray

21. 7 positions
2 alleles
2 strands
2 probes (PM/MM)
Total 56 features

23. Downstream Analysis (QC)

24. Current Practice: A Combination of Candidate Gene Approach and GWAS
Data-driven
Generates new knowledge
Relies on a SNP map
Hypothesis-driven
Constrained by knowledge
Allows systematic scanning
Candidate gene approach

25. Case Study: Genome-Wide Study of Exanta Hepatic Adverse Events
Ximelagatran, marketed as ExantaTM, developed by AZ
Developed/tested
Prevention of stroke in atrial fibrillation
Treatment of acute venous thromboembolism
Withdrawn from clinical development in 2006 because of ALT elevation:
Idiosyncratic nature: occurred in 6-7% of patients with ALT> 3 x upper limit normal (ULN)
Geographic dependent: high incidence in Northern Europe compared with Asia
Hypothesis: Genetic factors could be involved
Approaches: GWAS and candidate gene approaches

26. Samples (Subjects or Patients)
The original set (Training set)
248 subjects from 80 regions in Europe (Denmark, Finland, Germany, Noway, Poland, Sweden and the UK)
74 Cases = ALT elevation > 3 x ULN
132 Control = ALT elevation < 1 x ULN
39 Intermediate Control = ALT elevation >1 x ULN and <3 x ULN
An independent data set available late time
10 Cases and 16 Treated Controls

27. Experiment Design and Process
Candidate gene
Approach
GWAS
Genotyping
690 genes
26,613 SNPs
SNP/gene=40
Phase I
266,722 SNPs
Association analysis of SNPs with elevated ALT:
Matched and unmatched case-control analysis
Fisher’s Exact test, ANOVA, logistic regression analysis; Multiple testing correction (FDR)
Haplotype and linkage disequilibrium (LD) analysis
145 genes
76 genes
42,742 SNPs
SNP/gene=200
Phase II
28 SNPs
Representing 20 top-ranked genes

28. Drill-Down and Knowledge-Driven Analysis
Candidate gene
Approach
HLA-DRB1 region
HLA-DQA1 region
p-value SNP
A lowest
690 genes
26,613 SNPs
SNP/gene=40
Phase I
DRB1*07
Haplotype
145 genes
76 genes
DQB1*02
42,742 SNPs
SNP/gene=200
Phase II
28 SNPs

29. Validated by the Test Set
Test set (replication study)
10 Cases and 16 Controls
Both DRB1*07 and DQB1*02 are significant
Only 2 of 28 SNPs are significant, might be due to:
False positive in Phase I
Lack of power
A note:
Phases I and II genotyping using the Perlegen technology
Replication study using the TaqMan assay

30. Summary
Emphasis more on the candidate gene approach; candidate genes were selected from
Involved in MOA of Exanta
Associated with elevated liver enzyme (e.g., ALT)
Derived from preclinical studies for Exanta
Found to be genetically associated with adverse effects
Supported by the findings in Phase I
Some evidence obtained from the candidate gene approach (select 145 genes from among 690)
No evidence from GWAS (76 genes were selected)
Reflected in the drill-down approach
Focused on the gene/region with the lowest p-value SNP from the candidate gene approach; both SNPs identified this way are significant
2 out of 28 SNPs are significant from GWAS

31. My general impression
This study presents the evidence from a comparative analysis between two approaches
Knowledge-guided vs high-throughput screening
Hypothesis driven vs data driven
Less emphasis on GWAS and more reliance on the results from the candidate gene approach
Due to lack of power
Multiple testing correction issue
Is GWAS ready for the prime time?
Results from this study are not encouraging
Further investigation/survey is urgently needed