The map is based on more than a million single nucleotide polymorphisms (SNPs). SNPs generally arise by mutation and are then passed to the next generation. Recombination occasionally swaps portions of the DNA between two chromosomes in a pair during sexual reproduction, but in the absence of recombination two adjacent SNPs will be inherited together. A haplotype consists of a block of several SNPs along a stretch of the genome which is inherited as a unit. The Consortium examined the pattern of SNPs in 269 DNA samples from four different human populations and used the way in which adjacent SNPs occur together to build up the HapMap.

The real beauty of the HapMap, for researchers tracking down genetic components of disease, is that it makes it possible to select a small subset of SNPs which, because they almost always occur in blocks with other SNPs, can be used as proxies to screen large populations, looking for genetic differences. Many common diseases—such as cancers or cardiovascular disease—have a substantial genetic component, but not the simple one-to-one link between defective gene and disease that characterizes conditions such as cystic fibrosis or sickle cell anemia. Around 40% of the difference in susceptibility to these more complex diseases in a population is probably genetic. The rest is accounted for by a variety of other factors. The HapMap makes it possible to screen large populations in search of a relationship between particular haplotypes and the disease in question.

This approach has already borne fruit—for example, in isolating the genetic basis of age-related macular degeneration, a leading cause of loss of vision in the elderly, and the way it interacts with other genes and environmental factors such as smoking (see D.D.G. Despriet, et al., JAMA, 296[3]: 301-9, 19 July 2006). But while many more medical breakthroughs are certainly in the pipeline, the HapMap has also shed light onto fundamental biological processes.

It has, for example, pinpointed recombination hotspots. These are stretches where swapping is more likely to take place, often represented in the HapMap by discontinuities between haplotype blocks. In a detailed base-by-base study of part of the genome, the Consortium estimates that about 80% of all recombination events occur in about 15% of the sequence.

Also revealed is evidence of recent selection in the human genome. The HapMap showed a correlation between longer blocks and core biological processes, such as DNA repair and packaging. Genes that interact with the environment, such as immune-system genes, are associated with shorter than average blocks. "It is," as the authors note, "intriguing to speculate" that this reflects natural selection; long blocks suggest that the genes they contain have been conserved in much the same form in all the individuals studied, while shorter than average blocks suggest selection for greater genetic diversity.

There is considerably more direct evidence of selection too. People in some of the studied populations share some haplotypes not seen in the others. This suggests selection within that population. The gene that allows Europeans to digest dairy products is one such gene.

The HapMap sheds light too on the roughly 5% of the sequence that is highly conserved across all species so far studied. But only half of these super-conserved regions are in expressed genes. They could represent very important, albeit non-coding, sequences that have been conserved by so-called purifying selection, which gets rid of any mutants. Or, more prosaically, they could be recombination coldspots, preserved simply by not having been shuffled. The HapMap proves that they are not in fact coldspots but have been maintained by purifying selection, making them "of high interest for functional study."