360

26

Medicine and Disease

A further step in that direction would be taken by organizing clinical trials of pro-

posed new drugs such that patients are grouped according to their genetic profile.

Beyond that, the development of drugs tailored to haplotype seems feasible at first

sight, ²¹ especially with the introduction of microfluidics-based microreactors into

the pharmaceutical industry, which should make reliable small-scale syntheses eco-

nomically viable.

Undertaking gene therapy evidently requires knowledge of the genome. The pos-

sibilities of direct intervention at the level of the gene have been greatly expanded

by the discovery of small interfering RNA. Nevertheless, despite intensive efforts,

there has been no real success in the field to date. A major problem is the difficulty

of introducing the required nucleic acid material into cells from an external source.

Genome-wide association studies (GWAS) aim to scan entire (personal) genomes

in order to identify genes associated with certain diseases (phenotype), especially

polygenic ones. GWAS appear to have first been proposed by Risch and Merikangas

(1996). They observed that few of the numerous reports of genes or loci that might

underlie complex diseases have stood up to scrutiny. They analysed linkage analysis

and compared it with association analysis, using an unexceptionable model: The

disease susceptibility locus has two alleles, A and a, and the genotypic relative risks

(the increased chance that an individual succumbs to the disease) for genotypes aa,

aA and AA are assumed to be 1,gammaγ andgamma squaredγ², respectively. The association assumption

states that the more often affected siblings share the same allele at a particular site,

the more likely the site is close to the disease gene. The expected proportion of alleles

shared by a pair of affected siblings is ²²:

upper Y equals left parenthesis 1 plus w right parenthesis divided by left parenthesis 2 plus w right parenthesisY = (1 + w)/(2 + w)

(26.1)

where

w equals p q left parenthesis gamma minus 1 right parenthesis squared divided by left parenthesis p gamma plus q right parenthesis squaredw = pq(γ −1)²/(pγ + q)²

(26.2)

where pp and q equals 1 minus pq = 1 −p are the population frequencies of A and a, respectively. If

there is no linkage, upper Y equals 0.5Y = 0.5. For p equals 0.1p = 0.1 and gamma equals 4.0γ = 4.0, upper Y equalsY = almost 0.6 and slightly

less than 200 families would be required to make a reasonable inference of linkage.

On the other hand, for the probably more realistic values of p equals 0.01p = 0.01 and gamma equals 2.0γ = 2.0,

upper Y equalsY = 0.502 and almost 300,000 families would be required, which is practically

unachievable. Risch and Merikangas argue that association rather than linkage tests

enable an inference to be drawn from far fewer families.

Impetus in this direction came from the international HapMap project, which

was based on the sequencing technology developed for the human genome project

and which aimed to produce a genome-wide map of SNPs (Sect. 14.4.3). As Ter-

williger and Hiekkalinna (2006) have written: “The international HapMap project

was proposed in order to quantify linkage disequilibrium (LD) relationships among

human DNA polymorphisms in an assortment of populations, in order to facilitate

21 These developments are generally referred to as pharmacogenomics.

22 See Risch and Merikangas (1996) for more about the assumptions behind these formulae.