By Clare Sansom

The first human genome sequence – published in draft in 2000 and in a more complete version three years later – took well over ten years and cost $3-4 billion to complete. In the decade since that first draft launched the "human genome era", DNA sequencing technology has advanced to a point where companies like Illumina are offering to sequence an individual's genome for less than $10,000. NIH director Francis Collins has predicted that whole genome sequencing may become routine within little more than another decade. Yet the scientific community has only had a short time to consider the effect access to this genetic data will have on individuals and society. Many questions remain, including how environmental factors interact with the genome to influence the risk of disease, particularly cancer. The first personal genomics companies have now been accepting customers for more than two years, deriving health and ancestry information from analyses of up to a million of the most "informative" sequence positions of the three-billion-base genome. Recent controversies in the regulation of these companies show that the need to understand these complex genetic relationships is higher than ever.

Recently, I was offered the opportunity to have a selection points on my own genome analysed as part of a workshop on genomics at a conference organised by the Association of British Science Writers. I was particularly interested to see to what extent these Single Nucleotide Polymorphisms (SNPs) could predict my risk of various cancers. The kit allocated to me came from the US-based company 23andme[1]. I was asked to fill a small vial of saliva, seal it with a preservative and send it to the company by FedEx. About six weeks later the company posted my results on a password-protected area of its website.

Mark Henderson, science editor of The Times and author of 50 Genetics Ideas You Really Need to Know[2], submitted samples to three companies and had received results from two, 23andme and deCODEme[3], by the day of the meeting. This allowed a useful comparison to be made. In Henderson's case, the two reports showed discrepancies in only 77 positions out of over half-a-million that were sequenced by both companies, an error rate of fewer than one in 14,500.

My own data included reports of my genetic risk for eighty-two different diseases including seventeen types of cancer, each labelled with a confidence level. Clicking on the name of any of the "four-star" diseases, for which predictions are based on large, established research studies, leads the user to a resource describing the biology of the disease and how far the variants mapped determine risk,

The variants found in my genome revealed that I am at "slightly higher" risk than an average European of developing chronic lymphocytic leukaemia and thyroid cancer, at "slightly lower" risk of basal cell carcinoma and bladder cancer, and at average risk of other cancers. None of these, however, modified my risk by more than a few percent. A more potentially worrying result was a test for mutations in the BRCA 1 and 2 genes that can increase a carrier's lifetime risk of breast cancer to over 80%. Those results were located within a separate set of results labelled "carrier status" rather than "disease risk", and were locked so I could not read them without first opening a page of information about the test. Given my negative family history – only one possible case of hormone-dependent cancer, a great-aunt who developed what may have been ovarian cancer in her fifties – I was unsurprised to read that I carried none of the BRCA mutations tested for. Yet this is easy information to misinterpret, and so it can be dangerous. 23andme only extracts SNP data at three positions of relatively common mutations that are known to carry disease risk. Carriers of rarer mutations, who may be just as much at risk, will be missed. Had I any significant family history of the disease I would have been very reluctant to open my results.

People who are, or who believe themselves to be, at high risk of developing a particular cancer can often be offered genetic testing. Christine Patch, a genetic counsellor based at Guy's Hospital in London, works with women at high risk of breast cancer and families that carry rare disease-causing mutations. "We start by assessing a woman's risk of breast cancer from her family history alone", she says. "If it is relatively low we suggest routine breast screening, and if it is moderate we recommend mammography from the age of 40 rather than 50. Only women with breast cancer with a high-risk family history are offered BRCA1 and 2 gene screening for mutations, and if one is found other women at risk in the family would be offered testing". This screen is very reliable – much more so than 23andme's – but it is not perfect, as it may still miss some rare disease-causing variants.

Although an individual woman carrying a mutation in a BRCA gene is at an extremely high lifetime risk of breast cancer, these mutations are rare, accounting for only about 5% of all cases of the disease. There are other genes with much more common variants that are each known to increase breast cancer risk by a relatively small amount. Paul Pharoah, a genetic epidemiologist in the Department of Oncology at the University of Cambridge, UK, studies cancer risk prediction from common genetic variants. Working with colleagues across Cambridge, he has identified a group of seven genes that have variants associated with breast cancer[4]. Although the contribution of each of these to cancer risk is low, a woman carrying two copies of each of the high-frequency alleles has a lifetime risk of 23% of developing breast cancer, compared to an average of just over 9% in the general population. This is not sufficiently predictive to be of use to an individual woman, although it may be used to target mammography screening, increasing its availability to those women who may be at highest genetic risk. As yet, however, there are not enough common variations mapped to push many women even into the category regarded as "moderate" risk using their genetic profile alone. Unless and until many more variants have been mapped, accelerated mammography programmes will probably remain available only to women carrying BRCA mutations.

It is interesting that 23andme offers, besides the BRCA test, an analysis of common breast cancer risk alleles. It currently lists three of these, only one of which is definitely covered in Pharoah's paper. I was found to have one "low risk" and two "average risk" variants, which together reportedly decrease my lifetime risk to a percentage point or so below average. Yet, if or when 23andme examines the other six markers from this study, I may well find that I have more than enough high-risk variants to counterbalance this, pushing my overall risk higher than average.

This underlines one of the largest drawbacks of personal genomics: indeed, of any genetic testing. Unless or until every variant of every gene has been catalogued (which may well never happen) it will be possible for sequencing to miss significant risk or protective variants. Current genome scans are mainly limited to SNPs, which show only a small proportion of the whole story, and there is still much debate about the predictive value of the genome wide association studies that form the basis of most of the disease risk information. Once personal genomics routinely includes whole genome sequencing, capturing the greater depth and complexity of an individual's genome, disease risk statistics will still be a moving target. In addition, companies such as Oxford Nanopore are developing technologies that are expected to offer direct analysis of genome methylation, which is particularly important in oncology.

I would not advise anyone to sign up with 23andme without both a relatively clear family history and a reasonably good knowledge of genetics. And yet this is one of the most reputable companies, backing up the links it makes between genotype and disease risk with clear citations in the peer reviewed literature. In contrast, Pharoah cites one company in the US which has marketed a test for ovarian cancer risk based on a single point mutation in a kinase gene. "There is not even any evidence that this particular variation is associated with cancer", he says.

On balance, I am glad to have agreed to have my genome analysed. I have learned some useful things about my personal genetic inheritance – knowing that I am at moderately increased risk of developing age-related macular degeneration should lead me to visit the optician more regularly – and I have been relieved to learn that I appear not to be of particularly high risk for several cancers. Yet I am aware of the limitations of these studies, and how far they are from clinical utility. For the foreseeable future, let personal genomics remain a hobby for the curious, well-informed, and either moderately wealthy or well-connected. Those at real genetic risk will always need far more specialist advice and care.

[1] https://www.23andme.com/

[2] Quercus Publishing (2009). ISBN 978-1847246714

[3] http://www.decodeme.com/

[4] Pharoah, P.D.P. et al., N Engl J Med 2008; 358: 2796-803.