Investigators:
Daniel F. Heitjan, PhD (PI),
Wade H. Berrettini, MD, PhD, Caryn Lerman, PhD
Bayesian Analysis in Pharmacogenetic
Trials of Nicotine Dependence Treatment
Pharmacogenomic science
has
identified numerous genetic polymorphisms that influence patient
response to
drug treatment, yet it is believed that many more remain undiscovered. The search for important factors is
complicated by the large number of polymorphisms – potentially there
are
millions – of which the vast majority will have no function. Thus discovery approaches based on
conventional statistical methods, which as typically applied have a 5%
false
positive rate, may be too sensitive and insufficiently specific. Standard multiple-comparisons corrections may
go too far in the other direction, leading to procedures that are
exquisitely
specific but insufficiently sensitive.
Moreover, standard approaches are unable to make use of
potentially
relevant prior data, such as the situation of a SNP within a gene known
to be
related to drug metabolism or action.
Bayesian hypothesis testing methods are generally less sensitive
and
more specific than conventional tests, handle multiplicity more
gracefully, and
can make use of prior information. In
this application we propose to develop Bayesian methods for the
identification
of important SNPs in data from two clinical trials of nicotine
dependence
treatment.