The Genom of Homo sapiens.pdf

HUMAN SEQUENCE VARIATION 61Figure 4. LD analysis of Chromosome 20. Haplotype blocks (red and black boxes) were computed from LD data on Chromosome 20and are shown for two regions, one each of high and low overall LD. The analysis is repeated using data from different densities ofSNPs (average marker spacings in each analysis are listed on the left of the figure, and increasing SNP density is indicated by the verticalarrow).gion in high LD was estimated to be 65% in the North Europeansand 45% in the African-Americans (Ke et al.2004). The coverage estimates for North Europeans fitsthe trend in Table 1. However, the lower coverage figurefor the African-Americans suggests that more dense SNPsets may need to be used in these populations—echoingthe earlier findings of the Gabriel study (Gabriel et al.2002). The block patterns also varied significantly whenusing different analytical methods, indicating the need forother ways to describe LD patterns on a fine scale. For example,LD unit maps (Maniatis et al. 2002) reflect continuousviews of LD, although they are to some extentsensitive to varying allele frequency. A promising approachis the development of methods to estimate recombinationrates on a fine scale using the same data sets describedabove (McVean et al. 2004). These approachescould enable a precise assessment of progress in characterizationof LD in the genome, identify and prioritize regionsthat require more data, and help assess the ability ofpanels of anonymous markers to detect unknown diseaserelatedvariants.IMPLICATIONS FOR FUTURE STUDIESThe first global study of patterns of common variationin the human genome has crystallized in the form of theInternational HapMap Project (International HapMapConsortium 2003). The density of SNPs that are nowavailable in the public domain is likely to be sufficient forthis initial analysis of LD and common haplotypes, althoughsome targeted SNP discovery to fill gaps may benecessary. This project will produce a freely availablegenome-wide panel of tested SNPs that can be assessedfor use in association studies. It is intended to reduce theneed for researchers to search for their own SNPs, and toenable better searches for disease variants in genes, chromosomalregions, or ultimately, the whole genome usingthe indirect association approach. Evaluation of the parametersthat govern association studies are still required,and the results of pilot studies might help calibrate theHapMap and other studies to fit more precisely with theiranticipated applications. In parallel, it is also necessary toaddress the challenge of developing sufficiently largesample collections with accurate, accessible phenotypeinformation. Without this, even a perfect SNP panel thatcovers 100% of the genome will lack the power to establishan association between genotype and phenotype.We should continue sequence-based discovery of variantson a genome-wide basis; this would give us an unbiasedand properly quantified picture of the extent of commonsequence variation in the human population,replacing the current simulation-based estimates. Achievingthe first step at this level would benefit from upwardof 100 individual human genome sequences, a task thatrequires a greater than tenfold increase in sequence outputand is likely to demand the successful implementationof new technologies for cheap, accurate, high-throughputresequencing. Discovering rare variants (m.a.f. < 1%)would require improvements of a further one or two ordersof magnitude in sequencing. At present, therefore,pursuing rare variants remains in the domain of individualcandidate genes and may be best carried out in conjunctionwith a genetic disease study. Both candidategene and whole-genome sequencing will also be instrumentalin characterizing somatic variants that cause cancer,as demonstrated, for example, by Davies et al.(2002). Systematic pursuit of these areas of research willhelp us gain a view of the complete spectrum of variantallele frequencies that contributes to human disease.As we learn more about the full extent of functional sequencesin the genome, our knowledge of the functionallyimportant subset of variants will increase (for furtherdiscussion, see Bentley 2004). Developing a panel offunctional variants for all genes would increase the powerof association studies. However, we must be aware of theunknown; any association study based on the direct approachneeds to take account of the possibility that thetrue causative variant is not included in the study. It istherefore necessary to establish the underlying mechanismand explain the functional significance of the variant.Within the medical field, understanding the mechanismthat underlies disease will be the most effective wayto make progress in diagnosis, alleviation, and the effectivetreatment of disease.