REVIEWS of previous knowledge. (ii) Disease-causing mu- Genetic Mapping in Human Disease tations often cause major changes in encoded proteins. (iii) Loci typically harbor many disease- 1,2,3,4,5 1,2,5 1,6,7,8 David Altshuler, * Mark J. Daly, * Eric S. Lander * causing alleles, mostly rare in the population. (iv) Mendelian diseases often revealed great com- Genetic mapping provides a powerful approach to identify genes and biological processes plexity, such as locus heterogeneity, incomplete underlying any trait influenced by inheritance, including human diseases. We discuss the penetrance, and variable expressivity. intellectual foundations of genetic mapping of Mendelian and complex traits in humans, examine Geneticists were eager to apply genetic map- lessons emerging from linkage analysis of Mendelian diseases and genome-wide association ping to common diseases, which also show familial studies of common diseases, and discuss questions and challenges that lie ahead. clustering. Mendelian subtypes of common diseases [such as breast cancer (15), hypertension (16), and y the early 1900s, geneticists understood by Sturtevant for fruit flies in 1913 (1). Linkage diabetes (17)] were elucidated, but mutations in that Mendel’s laws of inheritance underlie analysis involves crosses between parents that vary these genes explained few cases in the population. Bthe transmission of genes in diploid orga- at a Mendelian trait and at many polymorphic In common forms of common disease, risk to re- nisms. They noted that some traits are inherited variants (“markers”); because of meiotic recom- latives is lower than in Mendelian cases, and linkage according to Mendel’s ratios, as a result of altera- bination, any marker showing correlated segre- studies with excellent power to detect a single causal tions in single genes, and they developed methods gation (“linkage”) with the trait must lie nearby gene yielded equivocal results. to map the genes responsible. They also recognized in the genome. These features were consistent with, but did that most naturally occurring trait variation, while In the 1970s, the ability to clone and sequence not prove, a polygenic model. The idea that com- showing strong correlation among relatives, involves DNA made it possible to tie genetic linkage maps in monly varying traits might be polygenic in nature the action of multiple genes and nongenetic factors. model organisms to the underlying DNA sequence, was offered by East in 1910 (18). By 1920, linkage Although it was clear that these insights applied and thereby to molecularly clone the genes respon- mapping was used to identify multiple unlinked to humans as much as to fruit flies, it took most of sible for any Mendelian trait solely on the basis of factors influencing truncate wings in Drosophila the century to turn these concepts into practical their genomic position (2, 3). Such studies typically (19), and Fisher had developed a mathematical tools for discovering genes contributing to human involved three steps: (i) identifying the locus respon- framework for relating Mendelian factors and diseases. Starting in the 1980s, the use of naturally sible through a genome-wide search; (ii) sequencing quantitative traits (20). In the late 1980s, linkage occurring DNA variation as markers to trace inher- the region in cases and controls to define causal mapping of complex traits was made feasible itance in families led to the discovery of thousands mutation(s); and (iii) studying the molecular and for experimental organisms through the use of of genes for rare Mendelian diseases. Despite great cellular functions of the genes discovered. So- genetic mapping in large crosses (21). But there hopes, the approach proved unsuccessful for com- called “positional cloning” became a mainstay was little success in humans. mon forms of human diseases—such as diabetes, of experimental genetics, identifying pathways Genetic association in populations. A possible heart disease, and cancer—that show complex in- that are crucial in development and physiology. path forward emerged from population genetics heritance in the general population. Linkage analysis in humans. For most of and genomics. Instead of mapping disease genes Over the past year, a new approach to genetic the 20th century, genome-wide linkage mapping by tracing transmission in families, one might mapping has yielded the first general progress was impractical in humans: Family sizes are small, localize them through association studies—that toward mapping loci that influence susceptibility crosses are not by design, and there were too is, comparisons of frequencies of genetic vari- to common human diseases. Still, most of the few classical genetic markers to systematically ants among affected and unaffected individuals. genes and mutations underlying these findings trace inheritance. Progress in identifying the genes Genetic association studies were not a new remain to be defined, let alone understood, and it contributing to human traits was initially limited idea. In the 1950s, such studies revealed correla- remains unclear how much of the heritability of to studies of biological candidates such as blood- tions between blood-group antigens and peptic common disease they explain. Below, we discuss type antigens (4) and hemoglobin b protein in ulcer disease (4); in the 1960s and 1970s, com- the intellectual foundations of genetic mapping, sickle-cell anemia (5). mon variation at the human leukocyte antigen examine emerging lessons, and discuss questions In 1980, Botstein and colleagues, building on (HLA) locus was associated with autoimmune and challenges that lie ahead. their use of DNA polymorphisms to study linkage and infectious diseases (22); and in the 1980s, in yeast (6) and the finding of DNA polymorphism apolipoprotein E was implicated in the etiology Genetic Mapping by Linkage at the globin locus in humans (7, 8), proposed the of Alzheimer’s disease (23). Still, only about a and Association use of naturally occurring DNA sequence poly- dozen extensively reproduced associations of com- Genetic mapping is the localization of genes un- morphisms as generic markers to create a human mon variants (outside the HLA locus) were iden- derlying phenotypes on the basis of correlation genetic map and systematically trace the trans- tified in the 20th century (24). with DNA variation, without the need for prior mission of chromosomal regions in families (9). A central problem was that association studies hypotheses about biological function. The sim- The feasibility of genetic mapping in humans was of candidate genes were a shot in the dark: They plest form, called linkage analysis, was conceived soon demonstrated with the localization of Hun- were limited to specific variants in biological can- tington disease in 1983 (10). A rudimentary ge- didate genes, each with a tiny a priori probability 1Broad Institute of Harvard and MIT, Cambridge, MA 02142, netic linkage map with ~400 DNA markers was of being disease-causing. Moreover, association USA. 2Center for Human Genetic Research and Department of generated by 1987 (11) and was fleshed out to studies were susceptible to false positives due to Medicine, Massachusetts General Hospital, Boston, MA 02114, 3 ~5000 markers by 1996 (12). Physical maps pro- population structure, because there was no way to USA. Department of Molecular Biology, Massachusetts General viding access to linked chromosomal regions were assess differences in the genetic background of Hospital, Boston, MA 02114, USA. 4Department of Genetics, Harvard Medical School, Boston, MA 02114, USA. 5Department developed by 1995 (13). With these tools, posi- cases and controls. Although many claims of as- of Medicine, Harvard Medical School, Boston, MA 02114, USA. tional cloning became possible in humans, and the sociations were published, the statistical support 6Department of Systems Biology, Harvard Medical School, Boston, number of disorders tied to a specific gene grew tended to be weak and few were subsequently 7 MA 02114, USA. Department of Biology, Massachusetts Institute from ~100 in the late 1980s to >2200 today (14). replicated (25). of Technology, Cambridge, MA 02139, USA. 8Whitehead Institute for Biomedical Research, Cambridge, MA 02142, USA. Several lessons emerged from studies of Men- In the mid-1990s, a systematic genome-wide delian disease genes: (i) The “candidate gene” approach to association studies was proposed *To whom correspondence should be addressed. E-mail: [email protected] (D.A.); mjdaly@chgr. approach was woefully inadequate; most disease (26–28): to develop a catalog of common human mgh.harvard.edu (M.J.D.); [email protected] (E.S.L.) genes were completely unsuspected on the basis genetic variants and test the variants for associa- www.sciencemag.org SCIENCE VOL 322 7 NOVEMBER 2008 881 REVIEWS tion to disease risk. The focus on common vari- answer is natural selection: Mutations that cause lization. Finally, disease-causing alleles could be ants as a mapping tool was a matter of practicality, strongly deleterious phenotypes—as most Men- maintained at high frequency if they were under grounded in population genetics. The human pop- delian diseases appear to be—are lost to purify- balancing selection, with disease burden offset by ulation has recently grown exponentially from a ing selection. But if deleterious mutations are a beneficial phenotype (as in sickle-cell disease small size. As predicted by classical theory (29), typically rare, how could common