Wind tests?
Why do you need a DNA test?
Maxim Struchalin, "Habr"
(See numerous references in the original article – VM.)
Over the past ten years, companies have appeared in Russia that, based on DNA analysis, inform the client about his personal predisposition to a large number of different diseases, inclinations to any sports or character traits. These companies give you an assessment of your personal risk, mainly using information from published scientific papers about which parts of DNA and how strongly associated with a disease or trait. However, today, for most multifactorial diseases and signs, geneticists are able to assess risks with limited accuracy. The main reason is that we still know quite a bit about how our DNA works. Below I will tell you about the accuracy of these tests, what you risk by passing them, and what the expert community thinks about all this.
We will talk about companies that give estimates for a wide range of diseases and signs: cancer, cardiovascular, mental disorders, sports, nutrition, etc. Such DNA tests in English-language literature are called "direct-to-consumer tests". You send the company a sample, for example, of saliva, from which DNA is extracted. Then this DNA is analyzed. After that, in your personal account on the company's website, you can see what your predispositions to various diseases are. Direct-to-consumer here means that there is no specialist between you and the company who would prescribe a test for medical or other reasons, explain the test results, form recommendations based on the test results. The service often includes consulting after testing, but this is done remotely.
The first such company called 23andme launched its service in California in 2007. Starting next year and for the next five years, the US Department of Health, represented by its agency called the Food and Drug Administration (FDA), regularly tried to get any data from 23andme confirming the reliability of genetic tests. Without receiving such information, in 2013, the FDA banned the company from selling genetic tests for all diseases to its customers in the United States. Only in 2017, the ban on testing ten diseases was lifted. Today, the company is testing for predisposition to 20+ diseases, 30+ signs, for the presence of a genetic risk of 40+ diseases and for the effectiveness of 3 drugs. After all the proceedings, a document appeared on the FDA website that describes the limitations of such tests. Below are excerpts from this document (highlighted in bold) with my comments and links to scientific papers to show how the expert community looks at this kind of service.
DNA tests do not use all genetic variants that affect multifactorial disease. Thus, the risk of illness that is reported to you by the company is not your actual full risk of getting sick (section I, paragraph A). The fact is that, despite the huge amount of money and efforts that have been spent by world science in the study of the genetic component of multifactorial diseases, our understanding of diseases is still quite limited, and, as a result, the predictive ability of risk assessment models is not very high. Geneticists still have to discover many unknown genetic variants, both protecting against diseases and increasing their risk. It is clear from publications in peer-reviewed scientific journals that most of the scientific community is skeptical about such DNA tests because of their low accuracy. Only in the last few years, with the advent of large genetic databases (such as UK Biobank), the idea began to sound cautiously in the world scientific literature that today we can start thinking about introducing risk assessment models using genetics into clinical practice.
For most diseases, the accuracy of prediction based on genetics varies from AUC=0.6 to AUC=0.7 (AUC is the criterion for the accuracy of the model; AUC=0.5 is the zero predictive ability of the model, AUC=1 is 100% accuracy of the model). Today, a doctor can diagnose and predict the development of many diseases much more accurately without using genetic information. For example, an article for 2020 shows that information about the age, gender, race, smoking, cholesterol level, blood pressure and diabetes status of a patient gives a much higher prediction accuracy (AUC=0.76) of coronary heart disease than only genetics (AUC=0.61).
Dozens and even hundreds and thousands of genetic variants are often responsible for multifactorial diseases, each of which has little effect on the likelihood of the disease. Today we know from the most extensive studies only about hundreds or thousands of them and only for a small number of diseases. For this reason, the accuracy of existing models is relatively low. There are, of course, monogenic diseases in which one mutation gives a very large effect and, as a rule, a very serious disease. Such diseases are rare and, as a rule, it is difficult not to notice them. If you have a monogenic disease, then there is a high probability that you already know about it and are in contact with a doctor.
Different companies use different numbers of genetic variants to assess risk. Thus, your risk of getting sick may vary depending on which company you take the test (section I, paragraph B). Scientific studies often say only that this genetic variant is most likely associated with the disease. Geneticists often simply do not know how exactly this segment of DNA increases or decreases the risk. Therefore, it is often difficult to determine which options are best used for risk assessment. Different companies have different criteria for constructing models: some include variants that have been repeatedly confirmed in many large studies, the second include those that have been found in more modest studies, and the third include all variants in the genome. The accuracy of the risk assessment depends on the approach.
Environmental factors can influence the development of the disease, except for genetic factors (section I, paragraph C). In genetics, there is such a parameter as the heritability coefficient. He talks about what proportion of the variance of a trait is described by genetic factors. A coefficient equal to zero indicates that only environmental factors (such as nutrition, ecology, smoking, etc.) are responsible for the trait. Such a disease can be influenced by changing only the external environment. A coefficient equal to 100% indicates that only genetic factors are responsible for the trait. A meta-analysis of most scientific publications over the past half century has shown that the heritability coefficient for 18,000 studied diseases is approximately 50%. This means that, on average, only 50% of your risk of getting sick is described by genetics. It is quite possible that you have "bad genes", but at the same time "good environmental factors". For example, you have a balanced diet, no bad habits, live in a place where the environment is good. Because of this, your risk of getting sick may be much lower than that of people with "good genes". Therefore, it is cheaper and more practical to first think about changing your environment than about DNA testing.
These tests are not designed to diagnose illnesses, inform you about your current state of health, or use them to make medical decisions, including whether or not to take medications and in what quantity (section I, paragraph F). All this can only be done by a doctor. Even if your test shows that there is a high risk of any disease, there is a possibility that this is a mistake. For example, a recent article describes several cases in which the test gave an erroneous result. In one of them, the test showed that a 12-year-old girl had a genetic variant associated with a disease called Ehlers-Danlos syndrome. The patient visited seven cardiologists, two gynecologists, an ophthalmologist, a gastroenterologist, six therapists, a medical geneticist, examined four other family members and carried out repeated genotyping. As a result, after a lot of effort, the disease was not confirmed. A 2020 study showed that the test may mistakenly show the presence of an average of five genetic variants associated with diseases.
In contrast to the case described above, many do not take any action after the test has shown a high risk of any disease. For example, in one study, most of those who received results with an increased risk of cancer did not make any attempts to change their diet, exercise or any other actions to reduce the risk of the disease.
After analyzing the risk calculation models used by Russian companies, I got the feeling that their accuracy is quite low. At least for some diseases and signs. For example, to calculate the risk of nicotine addiction, one of the Russian companies took one or two genetic variants from five genes: CHRNA5, NRXN1, VPS13A, NRXN1 and LOC102724740. While another company took 7 variants from only one CHRNA5 gene. And the last one
I used two variants (rs6474413 and rs7004381), which are very close in the genome (97 base pairs) and, accordingly, most likely one of them is uninformative due to strong mutual correlation. There is also a study for 2019, where 566 genetic variants responsible for nicotine addiction were found. For some reason, the companies didn't include them in the model. That is, we see that, firstly, the models of companies vary greatly, and secondly, the latest scientific results are ignored.
An important point is the data policy that the company adheres to and that they will adhere to in the future. It is known that one of the largest such companies in the United States, FamilyTreeDNA, shares the genetic data of its clients with the FBI. In 2018, a study found that 79% of DNA test users in the United States agree that the police should have access to their genetic data. Thus, if you have sent your DNA for testing, no matter to which company, sooner or later other organizations will have legal access to this data. I remember the movie "Gattaca", where a large corporation hires according to the genetic profile of a person. It is also possible that there will be those who want to use modern knowledge of genetics (and, accordingly, your genome) to revive the ideas of human breeding, which were once adopted in Nazi Germany.
Another problem is the security of data storage. We all see how many leaks come from organizations that seem to have better data protection systems. For example, it was reported that information about 100 thousand Sberbank customers was leaked. Anyone who has worked in such organizations knows perfectly well that it is almost impossible to establish barriers to such leaks. There will always be an employee who, out of revenge for the company, a desire to earn money, or simply by mistake, will merge the data to the side. It's only a matter of time before a black market for genetic data appears. Today we can learn quite a bit about a person by DNA, but after 10-20 years, I think the predictive ability of models will be high enough to tell a lot about your character, habits and diseases. Do you want others to know more about you than you do about yourself?
We know that 50% of our genome coincides with our parents and our children, 25% with our grandchildren, grandparents, 12.5% with our great–grandchildren, great-grandparents, etc. Thus, by sharing their genetic material with companies, a person automatically gives part of the information about the genome of their relatives. Agree, it is unpleasant to realize that someone, even your close relative, gives a stranger, for example, half of the digits from the PIN code of your bank card. Even half may be enough to solve the entire PIN code. The danger is also that another relative of yours can give the other half.
DNA testing for many unfamiliar with the field looks like a kind of high-tech product that can give you important information about your health. For some, the product will be useful, and maybe even save someone's life. If you have a monogenic disease and the doctor cannot determine what kind of disease it is, then it makes sense to undergo DNA screening for monogenic diseases. But it's still worth doing it in serious medical institutions. For most people, the non–specialized DNA test that companies offer is a dubious waste of money. Most likely, the test will not change anything in your life. If you make any decision based on such a DNA test, then you have a great chance of making the wrong decision. Most doctors, after viewing the test results, will most likely tell you what you already know well – lead a healthy lifestyle. Even if you have a really high genetic risk of a multifactorial disease, the best advice is to lead a healthier lifestyle and thereby reduce the risk.
If the article was useful and helped to save several tens of thousands of rubles, then I will be glad to thank you a little, which can be said here.
P.S. My name is Maxim Struchalin. I have been doing genomics for more than 10 years. He defended his PhD thesis in the Netherlands in 2013, worked in New Zealand for a company that uses DNA for animal breeding. Now I work with animal and human genetics. Including with a foreign company providing a DNA testing service. If you have already done the test and want to learn more about how much you can trust the results of an assessment of a certain disease or symptom, then write to me m.v.struchalin@gmail.com .
Portal "Eternal youth" http://vechnayamolodost.ru