New technologies for genotyping diallelic markers (SNPs) were recently developed that may be lower in cost, and more easily automated than microsatellite markers (STRPs). The reduction in genotyping costs resulting from such automation may significantly impact the overall cost of studies of complex traits, which generally require large sample sizes. Use of multiple SNPs in linkage analysis can recapture the linkage information otherwise lost with such markers. Here we derive a measure of the multilocus polymorphic information content (MPIC) in the context of link-age analysis for a cluster of SNPs, and we explore the characteristics of uniform vs. clustered SNP maps, relative to STRP maps. Issues addressed in comparing the map structures include the information content for clustered or single markers, and map accuracy. To be cost-effective, SNPs should have a common allele frequency between 0.5-0.75. No more than five loci per cluster are needed. Some linkage disequilibrium between loci in a cluster is tolerable. In the ideal case, a uniformly spaced SNP map is more cost-effective than one composed of clustered loci. However, the genotyping cost per marker for diallelic markers can be at most 60% of the genotyping cost per marker for STRPs. The consequences of using clustered vs. uniform SNP maps are considered in the context of map inaccuracy and use of multipoint vs. pairwise linkage analysis. Overall, when marker information, map accuracy, and flexibility of analysis are jointly considered, an optimal solution may be use of maps with 2-3 SNPs per cluster. Genet. Epidemiol. 22:205-220, 2002. (C) 2002 Wiley-Liss, Inc.
