TUESDAY
NOVEMBER 27, 2012
2:00-3:00 PM
Clinical Research Building
6th Floor, Room 692
1120 N.W. 14th Street
Miami, Florida
EDWIN S. IVERSEN, PH.D.
ASSOCIATE RESEARCH PROFESSOR
DEPARTMENT OF STATISTICAL SCIENCE
DUKE UNVERSITY
Functional Annotation Signatures as Prior Information in Genetic
Association Studies
We describe the development and application of a model for the
prior probability of phenotype-genotype association that incorporates
data from past association studies and publicly available annotation
data. The model takes the form of a binary regression of the
indicator of association on a set of annotation variables; we construct
an informative prior distribution on the coefficients in this
model that is informed by an analysis of SNPs that have been found
previously to be associated and are housed in the GWAS Catalog
(GC). To this end, we constructed a matched case--control study of
SNPs in which the cases are drawn from the GC and controls are
identified from the HapMap database, Release 27, Phases II and III
merged genotypes. The set of functional predictors we examined
includes measures that have been demonstrated to correlate with
the association status of SNPs in the GC and some whose utility in
this regard is speculative. As a result, we expect that only a fraction
of the annotation variables will contribute to predicting association.
We employed shrinkage priors to reflect this belief and used the normal-
exponential-gamma (NEG) distribution for its ability to heavily
penalize weakly determined coefficients and to weakly penalize
those that are well determined. We evaluate the out-of-sample performance
of the model and demonstrate its scalability as a prior distribution
for GWAS scale association testing.