Thanks to genetic testing, I now know that If it were biologically possible to have a baby with Mark Henderson, Science Editor of the Times, that baby would be certain to have wet earwax.
And he or she would definitely not have cystic fibrosis. Science!
This is all in aid of a session at the UK Conference of Science Journalists exploring the world of genetic testing, hosted by Mark, Daniel Macarthur from Genomes Unzipped and others. As part of the session, various journalists were offered the chance to get their genes tested for free by one of the three leading companies providing such services. I had a brief chat to Daniel about it, got his recommendations, and signed up. Four days later, a testing kit from 23andme arrived on my desk. I knew that 23andme had recently swapped some samples in a technical blunder but after reading Daniel’s blog, I was convinced that it was unlikely to happen again. If it did, I would enjoy finding out that I was secretly a black woman.
An hour later, I had delivered a dollop of my finest sputum into the tube they provided… and realised that I was only about a third of the way up to the fill-line. Doing this in the middle of the office was not a smart move. Ten further minutes later, and to a crescendo of laughter from my colleagues, the tube was full, sealed in a biohazard bag (I try not to take this as an indictment of my breath) and sealed in a Fed-ex envelope. Four weeks later, the results arrived. The whole process couldn’t have been simpler.
In fact, it was perhaps too easy. Signing up to the 23andme site, verifying the code on my testing kit and preparing the sample took little more than an hour. I had to read and agree to documents that reassured me about the privacy of my information and provide consent to analyse my samples. The same documents warn about the possible psychological consequences of finding out your data and the limtiations of the resulting information (more on these later; meanwhile, I’ve uploaded the full consent form to Posterous so you can see it for yourself). Nonetheless, I was well aware of these risks. I could have found out that I have substantially high odds of developing life-threatening diseases. I could have discovered that I’m not actually related to my parents. This is not a bottle one can re-cork.
But I grapple with issues of genetics and risk on a daily basis, both as a science writer and as part of my work at Cancer Research UK. I assumed that I would be sensible enough about my results to avoid any deterministic anxiety or any sense of false assurance. The results probably wouldn’t affect my behaviour in any material way and besides, my family history is (to my knowledge) largely free of any severe chronic illnesses. But to be honest, the biggest driver was an irrepressible curiosity. Mark Henderson said it best when he first took part in a similar test:
“Perhaps a few thousand of the 6.5 billion humans alive today have yet had the privilege of finding out so much information about the DNA that makes them unique. Here was my opportunity to become an early part of the genomic age.”
So how does it feel to be one of these pioneers? Well, as it happens, fun but a little disappointing.
23andme provides two basic services for its customers – information about health and about ancestry. I’ll leave the ancestry element aside from now (I’m genetically close to people from East Asia – who knew?) The health aspects are more interesting. They’re based on an analysis of 550,000 sites across my genome, which vary from person to person by a single DNA letter. These variations (called SNPs) are small but significant – they can affect an individual’s risk of disease, traits from eye colour to earwax texture, and response to various drugs and chemicals. By profiling the SNPs, 23andme tells its customers about everything from their risk of cancer to their ability to resist malaria.
I click on Disease Risk and I’m faced with a league table of future ills. Conditions have been arranged into three groups, depending on whether my risk is higher, lower or comparable to average. All of these risk estimates are based on research – on massive “genome-wide association studies” (GWAS) that identify SNPs behind specific traits. Each prediction has a confidence ranking, which reflects the size and number of studies that underlie the calculations. Importantly, each risk is presented in both absolute and relative terms, i.e. my actual risk of developing that condition, and my risk compared to other members of the population.
And here we hit the big snag. The “population” in this case generally refers to white Europeans or North Americans, because the vast majority of GWAS studies have been done in these regions. They are where the money to fund research and the willingness to sign up for it have been concentrated. This is a big deal because the link between SNPs and traits varies between ethnic groups.
As an example, the site tells me that I would have “slightly curlier hair on average” were I European. I’m not, so I don’t. More significantly, it also tells me that I have roughly double the average risk of prostate cancer. That’s certainly worth knowing about – it would mean that I have around a 1 in 3 chance of developing the disease within my lifetime, compared to the 1 in 6 odds that most men have. But again, this only applies to Europeans. In fact, if I set the analysis to ‘African’, I find that with exactly the same genes, I suddenly have a lower-than-average risk of prostate cancer. There’s no data for Asian populations. All in all, there are only a handful of results that I can rely on – I have an average risk of Type 2 diabetes (compared to the average Asian chap) but a slightly higher risk of rheumatoid arthritis.
I was aware of this before I signed up for the test, so my disappointment actually matches by expectations. However, things will change. Large genomic studies are underway in China and since 23andme regularly updates its information to account for new research, my reports should become a lot more interesting in time.
For the moment, there’s the odd useful nugget, veering from the useful to the humorous. I know that I am not a carrier for a variety of debilitating genetic diseases, from cystic fibrosis to Tay-Sachs disease. I’m told I have dry earwax, which I can confirm. I have “substantially higher odds of heroin addiction,” which I have no intention of confirming. Morbid fascination aside, there is a tremendous amount to play with on the site. If you had a mind to do so, you could interrogate your raw genomic data and scan your chromosomes SNP by SNP. You can even work out the odds of certain traits in a child between you and another 23andme user – hence the introductory quip to this post. The whole experience is presented through a seamless user interface and a permeating sense of fun.
This, perhaps, obscures more serious concerns. The ethnicity issue puts me in a somewhat enviable position, where results that could come as genuinely worrying can be easily brushed aside, leading to what Tom Whyntie called “the best version of Top Trumps EVER.” But not everyone would be in the same carefree position. Mark, for example, has a higher risk of glaucoma that would probably have terrified me, had it appeared in my results. Given that my entire career and virtually all of my leisure pursuits are heavily dependent on vision, I have a potent fear of eye diseases.
The same applies to neurodegenerative conditions, which led me to discover my risk for Parkinson’s disease with some trepidation (it’s average, since you ask). When you first see your report, most of your health risks are plain as day, but a few select pieces of genetic data come “locked”. To uncover them, you have to make an active choice. Your LRRK2 gene, which affects the risk of Parkinson’s is one of them. Your BRCA1/2 genes, which affect the risk of breast, ovarian and other cancers, are on the list too.
There is a good reason why these have been singled out for secrecy. Unlike the other variants that 23andme looks at, which are common in the population and have small effects on our health, mutations in LRRK2 and BRCA1 are rarer but more dramatic. Before unlocking the LRRK2 information, the site tells you “You are about to learn whether you have a relatively rare mutation in the LRRK2 gene that raises your lifetime Parkinson’s risk to more than 50% compared to the population average of between 1% and 2%.” It’s a potential game-changer.
This sort of information can also ripple out to affect people who never consented to have their genes tested in the first place, such as close family members. If I had a high-risk LRRK2 variant, chances are that some of my close relatives would too. Imagine, as an extreme situation, what would happen if I had an identical twin. The “locked” status is understandable but it is the equivalent of a big red button that says “Do not press”. How many people would resist the temptation, especially if they have been curious enough to sign up to the service? To 23andme’s credit, it uses the opportunity to make people fully aware of what they’re letting themselves in for. Amidst a lengthy statement, it tells you that not everyone with the high-risk version will develop Parkinson’s, nor does the low-risk version protect against the disease.
In fact, 23andme provides a consistently high standard of information. The original research papers that drive the risk predictions are all cited. Each trait or disease is accompanied by a lot of material – basic science, interviews with doctors, relevant posts by members of the online community, and even a timeline of relevant research. You also get practical tips for each disease, ranging from lifestyle advice to screening recommendations to contact details for relevant organisations. Judging by the cancer-related tips, these are all carefully crafted and based on decent evidence.
As mentioned, risks are presented in relative and absolute forms. This is important because a tripled risk may sound like a lot, but it’s less of a concern if it applies to a rare disease with a low background risk. You also get information in natural frequencies – an approach I like. For example, my report on rheumatoid arthritis tells me that in a group of 100 typical Asian men aged 35-79, 0.9 will get the disease. However, if all of them had my genes, 1.1 would get the disease. Nothing too much to worry about there.
But even here, there are potential problems with reliability. Mark Henderson has had his genes tested by three companies thus far – 23andme, DeCODEme and Pathway. In some cases, the three companies gave different results for the same stretch of DNA. In others, their reports gave different background risks of certain diseases. As Mark writes:
“If 23andMe is correct, then even though I have a high risk genetic variant, my overall risk of [glaucoma] remains low. If deCODEme is right, I have a one in three lifetime risk. And if Pathway is right, the population risk is low, but the peculiarities of my genome dramatically enhance my chances of getting it.”
These problems are inherent to the different companies, the tests they use and the information they provide. But further problems arise when that information enters the brains of the customers, where it often comes crashing against an inability to process information about risk. For all the careful phrases around risks and uncertainties, some folks are going to treat their report as nothing more than a sophisticated horoscope, given extra weight by the spectre of genetic determinism. For example, one brown-eyed community member was told that their genes predicted a high probability of having blue eyes with just a 1% chance of having brown eyes. They found the results “unsatisfying” and “confusing”. And this is a trivial example – the emotions that accompany a risk prediction for a severe disease must be even more potent, and especially so for people who aren’t well-versed in genetics.
Clearly, some will need more help than others at interpreting their results and that help will come at a premium. Not every jobbing health professional will have the knowledge to advise people about the implications of their genetic tests. Specialised genetic counsellors exist and 23andme will refer you to one, but you pay for the service out of your own wallet. This comes on top of the standard cost of the test, which would set the average customer back by $499.
I have mixed feelings about whether the cost is justified. The results are wonderful fodder for the curious and the confident but their limitations (especially for non-white ethnicities) prevent them from being of any obviously practical value to your health. This may change over time as more data flood in and the price starts to fall. But this doesn’t feel like a technology where there is a real benefit in being an early adopter.
Best possible result: “Mutant X-gene detected. Increased risk of superpowers, protecting world that hates/fears you”
Worst possible result: “Your ancestry is 50% Craig Venter”
Updated because I didn’t accurately reflect the contents of the consent statements, which I saved and are more informative than I remember, and on cystic fibrosis (see comment below)