ne of the dreams of genome analysis is personalized medicine; given an insight into the details of a patient’s DNA, a doctor could prescribe a specific and maximally effective therapy. Another is what might be called personalized prediction. From the DNA, it would be possible to divine just what diseases might lie in the future, and take steps to avoid them. Both came a little closer to reality with the publication in February 2005 of the paper that has stormed to the #1 spot in the Biology Top Ten.

A team from Perlegen Sciences Inc., of Mountain View, California, working with computer scientists associated with the University of California, has sequenced more than 1.58 million single-nucleotide polymorphisms (SNPs) in 71 individual Americans of African, European, or Han Chinese descent. SNPs are like spelling mistakes or, less prejudicially, variants in the DNA, where a single code letter differs between two individuals. Perhaps some 10 million SNPs have been identified and sequenced to date, but most have been explored in just a few individuals. The Perlegen team, led by David Cox, homed in on a subset of SNPs that were widely represented in their sample populations.

The International Haplotype Map (HapMap) Project has already published a similar picture of human genetic diversity based on fewer SNPs but more people. About 157,000 SNPs from nine people had been looked at by both HapMap and Perlegen. The company compared the two and showed that both datasets were extremely accurate, 99.6% identical. This is both important and reassuring, because the data are being used to associate specific genomes with disease patterns, and inaccurate sequences can easily throw researchers off the scent.

The vast majority of the SNPs were present in all three populations, which suggests that they have been around since before humans emerged from Africa. Some, about 18%, were so-called private SNPs, present in only one of the three, and the vast majority of those (75%) were in the African-American population. This, along with the greater variability of SNPs in African-Americans, lends further weight to the argument that all human populations originated in Africa.

A crucial feature of the Perlegen data is that the SNPs have been bundled together into bins, in which the presence of one SNP is highly correlated with the presence of other particular SNPs. Correlation of this sort, where two different SNPs are found together in an individual more often than expected, is known as linkage disequilibrium. It occurs because when a mutation arises it tends to carry flanking DNA sequences with it as it spreads through the population. Recombination, which shuffles the genome, breaks the links between neighboring sequences and reduces linkage disequilibrium. Blocks of linked SNPs might be evidence of evolution at work because selection will tend to keep sequences intact. Cox’s team found that SNPs in regions that coded for genes were more variable among the three populations than those in non-gene regions, and the same held true for coding SNPs compared to non-coding SNPs. This suggests that local selection is helping to maintain the pattern of SNPs. There was no similar evidence for the private SNPs, so probably there are no major evolutionary differences among the three populations.

Perlegen’s results have been publicly available to all at a company web site, no doubt accounting at least in part for the paper’s popularity. Researchers can reasonably quickly evaluate hypotheses linking diseases and genome patterns. In addition, the vast number of SNPs, and the correlations among them, means it is possible to start disentangling the connections and causal pathways underlying complex diseases with a poorly understood but obviously important genetic component. Knowing about a single genetic risk factor is not much use when it comes to tailoring either treatment or prevention. Being able to see most of the genetic factors associated with either the response to a particular therapy or, further in the future, with preventative measures, could enable personalized approaches without knowing in detail how the genetic differences lead to different outcomes.