Analysis of Association at Single Nucleotide Polymorphisms in the APOE Region Eden R. Martin,* , † ,1 John R. Gilbert,* , † Eric H. Lai,‡ John Riley,§ Allison R. Rogala,* , † Brandon D. Slotterbeck,* , † Catherine A. Sipe,* , † Janet M. Grubber,* , † Liling L. Warren,* , † P. Michael Conneally, ¶ Ann M. Saunders,* ,  Donald E. Schmechel,* ,  Ian Purvis,§ Margaret A. Pericak-Vance,* , † Allen D. Roses,* , ‡ and Jeffery M. Vance* * Department of Medicine, †Center for Human Genetics, and Bryan Alzheimer Disease Research Center, Duke University Medical Center, Durham, North Carolina 27710; ‡Glaxo Wellcome, Inc., Five Moore Drive, Research Triangle Park, North Carolina 27709; §Glaxo Wellcome, Inc., Greenford Road, Greenford, Middlesex UB6 OH3, London, United Kingdom; and ¶ Department of Medicine and Molecular Genetics, Indiana University School of Medicine, Indianapolis, Indiana 46202 Received August 30, 1999; accepted November 12, 1999 The discussion of the prospects of using a dense map of single nucleotide polymorphisms (SNPs) to identify disease genes with association analysis has been ex- tensive. However, there is little empiric evidence to support this strategy. To begin to examine the practi- cal issues surrounding this methodology, we identified 10 SNPs in the region immediately surrounding the apolipoprotein E locus (APOE), an established suscep- tibility gene for Alzheimer disease. Our goal was to examine patterns of allelic association to begin to in- vestigate the question of whether APOE could have been identified using SNPs. Our strongest evidence of association was at the 2 SNPs immediately flanking APOE. © 2000 Academic Press INTRODUCTION One of the major goals of the Human Genome Project (HGP) has been to identify a set of easily typed mark- ers spanning the genome (Collins and Galas, 1993). As a result, thousands of microsatellites have been mapped and are in widespread use in genetic studies. With the achievement of this aim, the newest endeavor of the HGP is to identify hundreds of thousands of single nucleotide polymorphisms (SNPs) across the ge- nome. It is anticipated that, over the next 3 years, at least 100,000 –200,000 SNPs will be identified by the HGP (Collins et al., 1998). Through a separate initia- tive, the SNP Consortium (TSC) has been formed with the goal of generating 300,000 SNPs in the next 2 years, with at least half of the SNPs mapped by mid- year 2001 (Marshall, 1999). Although SNPs will have generally lower heterozy- gosity and, therefore, be potentially less informative than microsatellite markers, they have several advan- tages. SNPs are more abundant than microsatellites; they are estimated to occur, on average, every 750 – 1000 bp (Kwok et al., 1996;Wang et al., 1998). Analysis with SNPs may also be advantageous since they gen- erally have a lower mutation rate than microsatellites. These properties make SNPs ideal for gene identifica- tion through association analysis since they increase the chance of finding a marker polymorphism in asso- ciation with a disease allele. Association analysis at- tempts to identify disease genes by looking for associ- ations between genetic marker alleles and a disease phenotype in case-control or family samples. Associa- tion studies can be more powerful for localizing genes than linkage analysis, particularly when the contribu- tion of these genes to the disease is small, as would be expected for complex diseases (Risch and Merikangas, 1996). For association analysis to be useful, however, a dense map of markers will be required since associa- tions can generally be found over only small distances, say no larger than 1 cM and often much smaller (Bod- mer, 1986; Jorde et al., 1993). Although genome-wide association studies using SNPs have been proposed (Risch and Merikangas, 1996), there are many difficulties to be addressed be- fore this strategy becomes feasible. A more immediate use of SNPs will be for association analysis in candi- date genes or regions. SNPs may be used to examine candidate genes that have been implicated through their biological function or because they lie in a region identified by linkage analysis. Even when no obvious candidate gene exists in a region of linkage, SNPs may be useful for narrowing the region of interest. Regions identified by linkage analysis are often quite large, especially for complex disorders, easily spanning 10 –20 cM or greater (Hauser and Boehnke, 1997). Since associations are found generally over much smaller distances, one strategy for identifying the dis- 1 To whom correspondence should be addressed at Duke Univer- sity Medical Center, Box 3468, Durham, NC 27710. Telephone: (919) 402-2553. Fax: (919) 401-0166. E-mail: emartin@chg.mc.duke.edu. Genomics 63, 7–12 (2000) doi:10.1006/geno.1999.6057, available online at http://www.idealibrary.com on 7 0888-7543/00 $35.00 Copyright © 2000 by Academic Press All rights of reproduction in any form reserved.