The origins of people's different responses to drug treatment lie in the genetic differences between them.

The Human Genome Project worked out the sequence of the three billion chemical 'letters' (nucleotides) of the human genome. What it produced, though, was an average sequence - and each of us differs slightly from this average.

Since the Human Genome Project was completed, research has focused on genetic variation between human beings and working out which variants are important to our health; which affect our likelihood of developing a particular disease or reacting badly to a drug.

Two types of variation are common in the genome:

• changes in single letters (so-called 'single nucleotide polymorphisms', or SNPs, known as 'snips')
• changes affecting blocks of DNA; someone may be missing the block entirely (a deletion), while others might have one, two, three or more copies. This is known as copy number variation.

SNPs are like changing a single letter in the metaphorical 'book of life', while copy number variation is like whole paragraphs or pages being lost or duplicated.

SNPs have been known about for a long time, but copy number variation is less well understood. It appears to be quite common - affecting around 12% of the genome - and has already been found to cause a variety of conditions.

Genetically speaking, humans are around 99.6% identical to one another. In the 0.4% of the genome that differs are genetic features that affect our susceptibility to disease or response to drugs. By analysing SNPs (or copy number variation), researchers can search for these features, which are called genetic markers. This has been made easier by advances in technology, which allow hundreds of thousands of SNPs to be analysed at the same time - this is known as high-throughput genotyping.

Guilt by association

In these experiments, researchers are looking for an association between a genetic marker and disease susceptibility (or adverse reaction). This means that a particular form of a particular marker is found more often in people with one type of reaction to a drug than it is in people with another type.

Because all humans are so similar genetically, people with a disease will share many genetic markers that are nothing to do with their disease. How is it possible to know which are important to the disease and which are just shared by chance?

The trick is to compare the genetic make-up of people who have a specific disease and those who do not. If a particular genetic marker is present in, 80 percent of patients, for example, but only 20 percent of a healthy population, it can be a sign that the marker is increasing the risk of disease.

To reduce the possibility of chance correlations, large groups of patients need to be studied. A common approach is to compare patients with unaffected people who are as similar as possible to the patient (the case-control approach). It is also valuable to analyse as many genetic markers as possible.