Genome-wide association study combining incomplete high resolution SNP data with sparse markers from a linkage scan. W-M. Chen, G.R. Abecasis. Center for statistical Genetics, Dept. of Biostatistics, Univ. of Michigan, Ann Arbor, MI.
   With millions of single nucleotide polymorphisms (SNPs) identified and characterized, genome-wide association studies are underway to identify susceptibility genes for complex traits and diseases. Given limited genotyping resources, we propose an approach that can produce large increases in power for genome wide association scans. We show that, by combining high-resolution SNP genotypes for just a few individuals with sparse marker data from a typical linkage scan, genotypes for many related individuals can be inferred with high accuracy, and power for the genome-wide association analysis can be substantially increased.
   First, our algorithm involves calculation of a probability distribution for each missing genotype in pedigree. We implemented this step using the Elston-Stewart algorithm, for larger pedigrees, and the Lander-Green algorithm, for smaller pedigrees. Next, observed genotypes and probability distributions for unobserved genotypes are combined in a rapid association test that characterizes association between genetic variants and quantitative phenotypes in families.
   We investigate the properties of the method in different family structures by simulation. We identify optimal family genotyping strategies for many different pedigree structures and show that in many cases it is sufficient to genotype <50% of the individuals in family to infer >90% of genotypes with near certainty. To illustrate our method, we carried out genome-wide association analysis for 27 gene expression phenotypes in 20 CEPH families, in which ~0.8 million SNPs are only genotyped in grandparents and parents, and a subset of 6,564 SNPs are genotyped in all 168 CEPH individuals. In addition to increasing evidence for association of 15 previously identified cis-acting associated alleles, our genotype inference algorithm allowed us to identify 4 novel cis-acting associated alleles that were missed when analysis was restricted to individuals genotyped by the HapMap project. Our genotype inference algorithm and the proposed association test are implemented in computer programs GHOST and Merlin.