Transcript Media:SRich072506

Study Design Discussion
The Ghost of Candidate Gene Past and
the Ghost of Genome-wide Association
Yet to Come
Stephen S. Rich, Ph.D.
Wake Forest University School of
Medicine
Major Issues for Association
Competing study designs
– Candidate genes
– Genome-wide
Analysis methods
Bioinformatics
Costs
Who to be genotyped and when
What phenotypes
Motivation and Justification
Two complementary motivations
– Immediate genome-wide association study
Identify novel regions
No prior knowledge (admission of ignorance)
Gene-gene interactions
Gene-environment interactions
– Immediate candidate gene evaluation
Assumed knowledge (admission of omniscience)
Gene-gene interactions
Gene-environment interactions
Claim: It will be as expensive to apply the 500K Affymetrix
technology as to evaluate candidate genes using custom SNP
chips – unless the latter is free
Issues Related to Analysis of
Association
Replication (internal) - corrects for multiple testing?
Expected number of true positives across genome
Both Candidate Gene and Genome-Wide Association
studies can use the Law of Large Numbers to identify
a meaningful proportion of variants for which you
have power
Power at an individual locus
Power across the genome
Can existing results in any study narrow the
hypothesis space?
Discussion Points
Design – best for efficiency and scientific value
– Candidate genes can be chosen and investigated for a fixed cost
(original CARE $8.5M)
– Genome-wide association study is feasible (original CARE $2.4M)
Technological advances
–
–
–
–
Cost equivalence
Timing and processing of samples
CARE candidate genes to target smaller list (700 genes/7000 SNPs)
CARE follow-up of GWA results (1000 genes/10000 SNPs)
Analytically, methods largely worked out
– Binning allele frequency results
– Best way to pick the winners for subsequent typing
– Integration of statistical results with databases
Integration with other NHLBI (NIH) GWA scans
Discussion