<img loading="lazy" decoding="async" class="alignright size-full wp-image-3542" title="kidneys" src="https://blog-api.23andme.com/wp-content/uploads/2009/05/kidneys.jpg" alt="" width="278" height="206" srcset="https://blogcms.23andme.com/wp-content/uploads/2009/05/kidneys.jpg 350w, https://blogcms.23andme.com/wp-content/uploads/2009/05/kidneys-300x223.jpg 300w" sizes="auto, (max-width: 278px) 100vw, 278px" />
Chronic kidney disease (CKD), characterized by the gradual loss of the kidneys’ filtering ability, currently affects about 10-13% of adults in the United States.   Patients suffering from the most severe form of the condition, end-stage renal disease, require dialysis or a kidney transplant to survive.
Rare mutations that cause <a href="https://www.23andme.com/topics/carrier/polycystic-kidney-disease" data-wplink-edit="true">kidney disease</a> have been identified, but finding common variations that impact susceptibility to CKD has been difficult.   Research published in 2009 showed that a variation in the gene that encodes the most common protein in urine could increase risk for CKD.   Now new research, published recently in the journal <a href="https://dx.doi.org/10.1371/journal.pgen.1001039" target="_blank" rel="noopener noreferrer">PLoS Genetics</a>, replicates this finding and shows that the variant may act by influencing how the kidneys adapt to age-related risk factors for kidney disease.
The new results, obtained by analyzing DNA from several thousand Icelandic and Dutch individuals, upheld the previous finding that each copy of the more common A version of in the UMOD gene, which encodes the Tamm-Horsfall protein, is associated with about 25% higher odds of CKD.
To further assess the effects of , the researchers turned their attention to measurements of <a href="http://www.nlm.nih.gov/medlineplus/ency/article/003475.htm" target="_blank" rel="noopener noreferrer">serum creatinine concentration (SCr)</a>, an indicator of kidney function.   Higher SCr is associated with decreased filtering ability in the kidney.   They found that in the Icelandic sample,  the variant affects SCr levels in an age-dependent manner.   Before age 50, this variant doesn&#8217;t have much of an affect.   But thereafter each copy of a A is associated with higher SCr every year, especially around age 70.   The effect of was intensified in people with hypertension compared to those without a diagnosis of high blood pressure, as well as in those with type 2 diabetes compared to those without.
Based on their results, the researchers speculate that affects the vulnerability of the kidney to damage as a person ages.
&#8220;The common diseases happen at the interface between genes and the environment, and this study offers a clear and medically useful example of this dynamic in action.   This SNP is now a validated risk factor for kidney disease, but we have also shown it can be even more meaningful if looked at in the context of age and broader health history,&#8221; said the study&#8217;s senior author, Kari Stefansson, in a <a href="http://www.prnewswire.com/news-releases/decode-shows-how-genetic-risk-of-kidney-disease-frames-response-to-environmental-risk-over-time-99579559.html" target="_blank" rel="noopener noreferrer">press release</a>.

23andMe Blog

23andMe Blog Team

New Research Sheds More Light on Role of Genetic Variant in Chronic Kidney Disease

<img loading="lazy" decoding="async" class="alignright wp-image-5157 size-medium" title="migraine" src="http://35.88.189.111/wp-content/uploads/2009/10/migraine-300x214.jpg" alt="" width="300" height="214" srcset="https://blogcms.23andme.com/wp-content/uploads/2009/10/migraine-300x214.jpg 300w, https://blogcms.23andme.com/wp-content/uploads/2009/10/migraine.jpg 435w" sizes="auto, (max-width: 300px) 100vw, 300px" />
Headaches are a pain &#8212; literally. But while most people see them as a temporary nuisance, many others find them extremely debilitating. For the 8% of men and 17% of women who suffer from migraine headaches, these long, recurring episodes can include more than just throbbing pain: vomiting, chills, sensitivity to light or sound, or trouble concentrating are just a few of the possible symptoms. Some people experience a set of symptoms prior to their migraine attacks known as “aura” but for others the attacks come without warning.
Because the condition is complex and manifests differently in each person, the biological causes of migraine have been difficult to determine. New research, however, is implicating common genetic variants which may help to elucidate the biological underpinnings of the disease.
In a study published this week in <a href="https://dx.doi.org/10.1038/ng.652" target="_blank" rel="noopener">Nature Genetics</a>, a team of scientists from the International Headache Genetics Consortium (IHGC)  identified a genetic variant associated with migraine susceptibility. The researchers compared almost 6,000 migraine sufferers to more than 50,000 people who did not suffer from migraines, all of European ancestry, and found that the less common variant was associated with about 1.2 times higher odds of migraine headache.
Although the SNP is not located within any known genes, it is found in a region of the genome that contains genes involved in the regulation of a neurotransmitter called glutamate. When the IHGC team looked at the nearby genes, they observed that was significantly correlated with the expression levels of MTDH. MTDH regulates another gene that encodes a protein that transports glutamate in the brain. Previous research has suggested that an imbalance in glutamate levels may contribute to the development of migraines, in addition to other neurological disorders, but the exact mechanisms are still unknown. More studies are needed to confirm and elaborate on the IHGC’s findings.

Researchers Link Migraine Susceptibility to Genes Involved in Glutamate Regulation

<img loading="lazy" decoding="async" class="alignright size-medium wp-image-7164" title="dnablocks(iStock)" src="https://blog-api.23andme.com/wp-content/uploads/2010/08/dnablocksiStock-300x197.jpg" alt="" width="300" height="197" srcset="https://blogcms.23andme.com/wp-content/uploads/2010/08/dnablocksiStock-300x197.jpg 300w, https://blogcms.23andme.com/wp-content/uploads/2010/08/dnablocksiStock.jpg 427w" sizes="auto, (max-width: 300px) 100vw, 300px" />
Having been a genetics educator for over two decades, I was excited to learn in May that Berkeley’s incoming freshman class would have an opportunity to discover whether they have an AA, AG, or GG at a specific site near the LCT gene, along with similar information for the ALDH2 and MTHFR genes. &nbsp; Time and again I have seen students, friends and relatives pale at the mention of genetics concepts, until they get their own data and are instead forthcoming with questions and eager to learn. So the idea of university students learning about genetics through a small set of personal genetic data seemed an excellent one. The data would provide an opportunity for discussion of a broad range of topics, from genetics to concepts of race to the ethical issues surrounding medical genetics. And the organizers had thought long and hard about the project, developing, for instance, an <a href="http://webcast.berkeley.edu/media/common/media/one_off/pa_20100708_ls_h264fs.mov" target="_blank" rel="noopener noreferrer">online video consent form</a>.
Last week I was deeply disappointed to learn that the course organizers, in response to a decision by the California Department of Public Health, will not be providing students with their personal genetic data. Instead, they will provide aggregate data — summaries of how many students have each of these “genotypes” (for instance, AA, AG, or GG). &nbsp; I understand the importance of making sure that students’ privacy is protected, and that students are adequately informed before obtaining their data. And I think the topic of whether the CDPH decision makes sense is a question worthy of further discussion. &nbsp; But today I will focus on a different, related topic, because, after a brief period of disappointment over this decision, I began to think more constructively. From an educator’s perspective, there are still valuable genetics lessons to be learned via this initiative, even if students don’t gain access to their personal genetic data.
In fact, were I coordinating the project, even if there were an opportunity to provide students with personal data, I would refrain from giving students immediate access. &nbsp; Indeed, I would wait to provide access to aggregate data as well. &nbsp; I would start by inviting students to make an educated guess regarding their genotype for each of the genetic markers in the LCT, ALDH2, and MTHFR genes. &nbsp; Presumably the CDPH doesn’t have jurisdiction over students guessing their genotype, even though a guessing exercise would lead to far less accurate personal data than genotyping would. &nbsp; So how would students guess their genotype? &nbsp; Let’s take each of the markers one-by-one.
Lactose tolerance (LCT gene): A genotype of GG at one particular position near the LCT gene can lead a person to have trouble digesting the lactose in cow’s milk. &nbsp; Given that Berkeley students can drink milk legally, making a guess regarding their genotype (AA, AG, GG) for this marker is more straightforward than for the other two genes. They can base their guess regarding their “lactose tolerance” genotype on their own experience drinking milk, and the experiences of their family members. &nbsp; If, for example, they don’t digest milk well, and neither of their parents drinks milk comfortably, they might guess that their own genotype is GG. &nbsp; If only one parent has enjoyed drinking milk as an adult, their initial guess might be AG. And if everyone in their family drinks gallons of milk, their best guess would be AA. &nbsp; Next, I would provide them with frequencies of each of the LCT variants (A and G) in a number of populations. &nbsp; Such data are <a href="https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1234910/table/TB2/">readily available</a>. &nbsp; Students would be invited to put these two sets of data (personal experience, family history and population frequencies) together and guess their genotype, recording it for themselves and submitting it (anonymously) to a course instructor.
Alcohol metabolism (ALDH2 gene): A genotype of AA at one particular position in the ALDH2 gene can lead a person to have trouble tolerating alcohol so that they experience what is called <a href="https://www.23andme.com/topics/wellness/alcohol-flush-reaction">“alcohol flush”</a> and other discomforts. The steps for guessing one’s genotype are similar to those for LCT, except that most Berkeley freshman cannot drink alcohol legally in California. Therefore their guess regarding this marker will need to be based on information from parents or other older relatives. &nbsp; Students could find out if their relatives ever “flushed” after drinking alcohol. &nbsp; If two parents said yes, then the student might guess that they have the AA genotype for this locus. If the flushing is moderate, the student might guess that they are most likely to have the AG genotype. Next, I would provide them with frequencies of each of the ALDH2 variants (A and G) in a number of populations. As for LCT, students would be invited to put these two sets of data (family history and population frequencies) together and guess their genotype, recording it for themselves and submitting it (anonymously) to a course instructor.
Folic acid requirement (MTHFR gene): A genotype at one particular position in the <a href="http://circ.ahajournals.org/cgi/content/full/111/19/e289" target="_blank" rel="noopener noreferrer">MTHFR gene</a> can influence a person’s requirement for folic acid (found in, for instance, leafy green vegetables). Here most students would have little personal experience or family history to help them out. &nbsp; So they would need to rely on the population frequencies. &nbsp; They might even use a random number generator to help them make an educated guess. &nbsp; If one parent were affiliated with one population and the other parent with a second population, the student would use the relevant population frequencies of the two populations. &nbsp; As before, they would record their genotype guess for themselves and submit it to their instructor.
During this “guessing” process students could discuss how comfortable they are sharing the information with one another. &nbsp; Would they be comfortable making the information public? Do they see any differences between sharing their guesses and sharing their actual genotype data (were they to have access to it)? &nbsp; They might also discuss the implications of the different frequencies in the different populations. &nbsp; Is any variant at 100% in any population and 0% in another so that ethnicity predicts genotype? &nbsp; Are their stereotypes around the phenotypes of alcohol flush and lactose tolerance? What about the health implications of the information?
Once the students have submitted their guesses for all three markers, instructors would tabulate and report back the aggregate data. &nbsp; If students have also submitted information regarding their sex and ethnicity, data could be summarized within those categories. &nbsp; At this point it would be time to reveal the actual aggregate data to the students. &nbsp; The big questions would be: &nbsp; How close are the students’ guesses (in aggregate) to the actual data? &nbsp; Are they close for all three markers? If not close, what are the possible explanations? &nbsp; Lack of true correlation between genotype and the trait? &nbsp; Incorrect published estimates of variant frequencies? &nbsp; Laboratory error? Incorrect calculations by the students when they made their educated guesses? What else might be going on?
Under the current plan, this comparison of the summary of educated guesses and the summary of actual data would be the final chapter of the genetics component of the course. &nbsp; But maybe next year would be different. &nbsp; Maybe this year’s freshman class at Berkeley would pave the way for future classes to have one more small but highly significant step: &nbsp; comparing their individual educated guesses with their actual personal genotype data. &nbsp;The personal data would provide opportunities for several powerful lessons. &nbsp; For instance, a number of students would discover that they have variants that are often reported as being found only in ethnic groups other than their own, revealing the fallacy of equating race or ethnicity with biology.
Having been through the preparatory lessons and exercises, interested students would be prepared to gain access to their personal genetic information. And thousands of individuals would be better prepared to consider the genetic information that will be broadly available in the near future.

Lessons at Berkeley: Genetics and Beyond

<img loading="lazy" decoding="async" class="alignright size-full wp-image-3624" title="liver" src="https://blog-api.23andme.com/wp-content/uploads/2009/05/liver.jpg" alt="" width="309" height="229" srcset="https://blogcms.23andme.com/wp-content/uploads/2009/05/liver.jpg 350w, https://blogcms.23andme.com/wp-content/uploads/2009/05/liver-300x223.jpg 300w" sizes="auto, (max-width: 309px) 100vw, 309px" />
Hepatocellular Carcinoma (HCC) is the <a href="http://www.cancer.gov/cancertopics/pdq/screening/hepatocellular/HealthProfessional/allpages" target="_blank" rel="noopener noreferrer">fourth most common cancer</a> in the world.  More than half a million new cases of this form of <a href="http://www.mayoclinic.com/health/liver-cancer/DS00399" target="_blank" rel="noopener noreferrer">liver cancer</a> are reported each year.   Excessive alcohol consumption and obesity have long been linked to HCC, but the main cause of the disease is chronic <a href="http://www.mayoclinic.com/health/hepatitis-b/DS00398" target="_blank" rel="noopener noreferrer">hepatitis B virus</a> (HBV) infection.  Not everyone chronically infected with the virus, however, develops HCC.
To investigate whether genetic factors play a part in whether an HBV-infected person will go on to develop HCC, researchers analyzed the DNA of several groups of people with Chinese ancestry, as it is known that in China 80% of cases of HCC are linked to chronic HBV infection.  Their results, published online this week in the journal <a href="https://dx.doi.org/10.1038/ng.638" target="_blank" rel="noopener noreferrer">Nature Genetics</a>, show a strong link to a genetic variation on chromosome 1.
The sample of chronic HBV carriers used to identify the association consisted of 355 people with HCC and 360 without HCC.  An additional 1,962 individuals with HCC, 1,430 control subjects, and 159 families, were also studied to confirm the results.   Researchers found a genetic association for the condition.
The researchers also explored differences in the effect of on HCC rates among HBV carriers of different genders and ages.  They found that was more strongly associated with HCC in virus-infected males than females.   The variants were found near several genes: KIF1B, UBE4B and PGD.  Multiple lines of evidence suggest that one or more of these genes are plausible candidates for HCC susceptibility.  Changes in the region of the genome where they are found are commonly seen in many different cancers, including HCC.  The KIF1B gene has been shown to be important in preventing some types of tumors, and mutations in this gene are associated with several types of cancer.  UBE4B and PGD encode proteins involved in vital cellular processes that often go awry in cancer development.

Researchers Identify Genetic Variant That May Affect Liver Cancer Risk in Those Chronically Infected with Hepatitis B Virus

This week, Mary Carmichael of Newsweek is exploring the world of direct-to-consumer (DTC) genetic testing and deciding whether or not she&#8217;d like to find out about her own DNA information.
<img loading="lazy" decoding="async" class="aligncenter wp-image-6586 size-full" title="23andMe_Logo_blog" src="https://blog-api.23andme.com/wp-content/uploads/2010/05/23andMe_Logo_blog.jpg" alt="" width="350" height="349" srcset="https://blogcms.23andme.com/wp-content/uploads/2010/05/23andMe_Logo_blog.jpg 350w, https://blogcms.23andme.com/wp-content/uploads/2010/05/23andMe_Logo_blog-150x150.jpg 150w, https://blogcms.23andme.com/wp-content/uploads/2010/05/23andMe_Logo_blog-300x300.jpg 300w, https://blogcms.23andme.com/wp-content/uploads/2010/05/23andMe_Logo_blog-271x271.jpg 271w" sizes="auto, (max-width: 350px) 100vw, 350px" />
23andMe was asked to contribute a short piece on what you can–and can&#8217;t–learn from services such as ours. We were more than happy to help, as it is our firm belief that education is a critical first step toward our goal of meaningful integration of genetic information into clinical care. Below is an expanded version of <a href="http://www.newsweek.com/2010/08/03/dna-dilemma-day-two-what-can-i-learn-from-at-home-dna-tests.html" target="_blank" rel="noopener">what we wrote</a>.
If you&#8217;d like to join the conversation, leave a comment over at Newsweek, or <a href="https://twitter.com/mary_carmichael" target="_blank" rel="noopener">follow Mary on Twitter</a> and let her know what you think.
<h6>Get the Right Drug at the Right Dose</h6>
What we can tell you: Genetics can impact whether you&#8217;ll need more or less of a medication compared to most people, or whether you might be at increased risk for serious side effects.
For example, the blood thinner warfarin (Coumadin ®) can help prevent blood clots, but it can also cause excessive bleeding in people whose sensitivity to this drug is increased by two genetic variations covered by 23andMe. There are also genetic effects on clopidogrel (Plavix ®) efficacy.   Normally this drug helps prevent heart attacks by keeping blood cells from sticking together. But a genetic variation that interferes with the drug&#8217;s metabolism prevents some people from getting the full effect, putting them at increased risk for cardiovascular events.
What we can&#8217;t do: DTC genetic testing can&#8217;t replace your doctor when it comes to making decisions about how best to treat you.   There are many factors other than genetics that need to be taken into consideration.   No one should stop taking a medication or change their dosage without a doctor&#8217;s supervision.
<h6>The Future: Mutations that Could Impact Your Children</h6>
What you&#8217;ll learn: If two people are carriers of certain disease-causing mutations, each child they have together will have a 25% chance of being born with the disease.  Knowing a child has a chance of being born with one of these diseases can facilitate early intervention and improve outcomes.
One in 29 people of European descent carries a mutation that can cause cystic fibrosis.   Approximately one in five Ashkenazi Jews carries a mutation associated with a severe disease that could result in the early death of a child, such as Tay-Sachs, Canavan disease or Niemann-Pick disease type A. Sickle cell anemia occurs in about one out of 625 African Americans and one out of 1,000 Hispanic Americans.
23andMe tests for mutations associated with these and many other inherited conditions.
But remember: For almost every inherited condition, there are many, many known mutations.   Some of these are more frequent than others, and the chance of having certain mutations often varies by ethnicity.   All screening tests necessarily cover only a subset of the possible mutations for each disease.   Because of this, these tests can tell you that there is a chance that a child will be affected, but if no mutations are detected you don’t know that a child absolutely won’t be affected.   The presence of undetected mutations in each parent could still lead to a child being born with the disease.
<h6>Common Diseases — the Future of Genetics</h6>
What we know: Common DNA variations found within all of us have been linked to increased risk for things like diabetes, cancer and autoimmune conditions.Learning your genetic risk for these conditions allows you and your doctor to focus on what matters most for you.
What&#8217;s left to figure out: The genetic variations identified to date explain only a small part of a person&#8217;s risk for each disease.   Unknown genetic factors, as well as environment, are also important.   Additionally, because most research is conducted in populations with European ancestry, people from ethnic minorities get limited information from DTC genetic testing.
Through its research, 23andMe is working to improve the information that all people can get from their DNA data.   Our 23andWe surveys allow everyone to contribute to work aimed at understanding some of the most basic facets of our biology, and have already resulted in the <a href="https://blog.23andme.com/2010/06/24/plos-genetics-23andme/" target="_blank" rel="noopener">publication of a report in a peer-reviewed scientific journal</a>.   Our Parkinson&#8217;s disease and sarcoma communities are bringing people from around the world together in a participant-driven effort to make real strides in our understanding of these diseases.
We are also partnering with several groups in order to increase the diversity in our database and, in turn, the number of results we will be able to provide to all people, regardless of their ancestry.
<h6>The Fun Stuff</h6>
Some of the information hidden in your DNA is just plain cool!
Physical Traits: Is your earwax wet or dry?   Never thought about it?   Well, you will once you learn about how just one little letter in your DNA can make all the difference.   Your eye color, hair curliness, whether your sneeze upon entering bright light, and the ability to smell asparagus in your urine (ewwwwww!) are also all influenced by genetic factors.
Ancestry: Locked within your genetic code are both the story of your ancestors and the information you need to fill in your family tree.
Ancient Migrations: With state-of-the-art tools 23andMe traces maternal ancestry back tens of thousands of years with a small piece of DNA passed down from mother to child. Males can uncover their paternal ancestry through the Y chromosome that is passed down from father to son.
Global Origins: By comparing your DNA to reference populations from around the world, 23andMe can give you clues as to where each piece of your genome originated and which populations you are most similar to.
Find Relatives You Never Knew You Had: 23andMe&#8217;s Relative Finder lets you find all the other 23andMe members who match your DNA. You can browse and make contact anonymously.   Most people find over 100 &#8220;cousins,&#8221; though this varies based on ancestry.   People who have mainly European roots may find several hundred relatives. If you have Ashkenazi Jewish ancestry, you might find well over a thousand.

What DTC Genetic Testing Can–and Can’t–Tell You

As we posted here on the Blog yesterday, the Subcommittee on Oversight and Investigations of the House of Representatives Committee on Energy and Commerce held a hearing on “Direct-To-Consumer Genetic Testing and the Consequences to the Public Health.”
Central to this hearing was an investigation by the Government Accountability Office (GAO) into 15 DTC genetic testing companies.
The GAO refused to discuss its concerns or its report with 23andMe, and now that the report is public and we have had a chance to review it, we are troubled and find the report is deeply flawed. We note that while such an exercise as conducted by GAO has the potential to raise questions, it does not provide the answers that a more rigorous scientific study would provide.   This report raises questions, but leads to few conclusions because of its unscientific nature.   The GAO itself recognizes this, writing, &#8220;It is important to emphasize that we did not conduct a rigorous scientific study.…&#8221;
[The report is accessible from the <a href="https://energycommerce.house.gov/index.php?option=com_content&#038;view=article&#038;id=2083:hearing-on-direct-to-consumer-genetic-testing-and-the-consequences-to-the-public-health&#038;catid=133:subcommittee-on-oversight-and-investigations&#038;Itemid=73" target="_blank" rel="noopener">Committee&#8217;s website</a> (Click on PDF for &#8220;Testimony of Gregory Kutz.&#8221;   Gregory Kutz is Managing Director, Forensic Audits and Special Investigations, Government Accountability Office).]
 
We are confident in our service&#8217;s accuracy and reliability. It is widely accepted that the technology we are using is sound.   We understand that GAO did not find any problem with the underlying data that we provide — the As, Cs, Ts and Gs.   What is at question is whether or not one part of the information about that data that we provide is of value, and we believe strongly that it is.
The GAO report focused only on disease risk probabilities.   It did not focus on ancestry or the trait reports we offer.   It also failed to address that we also provide information about carrier status for single gene diseases such as cystic fibrosis and Tay-Sachs disease, as well as the fact that we provide information about a customer&#8217;s likely response to certain prescription medications that have been shown in clinical trials to have differing effects and side effects depending on a person&#8217;s genetic make-up.   This suggests that GAO found no problems with these parts of our service.
Carrier status and drug response information are clearly useful. In fact, Dr. James Evans, the Director of Adult Genetics Services at the University of North Carolina and the Editor-in-Chief of Genetics in Medicine, admitted during the Congressional hearing that drug response information would be of great interest to him as a physician.   (He was specifically referring to results pertaining to a patient&#8217;s sensitivity to the anti-viral medication abacavir.)   It should be noted that during the hearing it was not clear that Dr. Evans had been the primary consultant to GAO regarding the scientific and medical relevance of the results provided by DTC genetics testing companies.
The remarks made by<a href="https://blog.23andme.com/2010/07/20/23andme-at-fdacdrh-public-meeting-on-oversight-of-laboratory-developed-tests-ldts/" target="_blank" rel="noopener"> 23andMe Co-Founder Anne Wojcicki and General Counsel Ashley Gould</a> at the FDA public meeting on July 20, 2010 about laboratory developed tests demonstrate the importance of the work we are doing and our commitment to ensuring that members of the public are provided unfettered access to their DNA information in a responsible manner.   We embrace the ideas that the FDA offered today about stepping in to provide a regulatory framework and help set scientific and transparency standards across the industry.   We look forward to helping with this process.
Read on for discussion of some of the problems with the GAO report.
One of the most unfortunate parts of the GAO report is that it unfairly lumps together reputable and well known companies such as 23andMe with un-named companies making verifiably untrue endorsement claims, spurious scientific claims, and also selling potentially fraudulent supplements in addition to genetic testing services. Some of the most troubling of these interactions between the GAO and genetic testing companies can be found in a table on pages 15-16 of the GAO report, and in a Youtube video the Office has posted.
It must be noted however, that although the companies are not identified in the video or the report, at the hearing it was revealed that 23andMe is Company 1.   Other than saying that we believe customers should consult with their physician or other healthcare professional when they have questions about their data, 23andMe/Company 1 is not implicated in any wrongdoing.
GAO seems to believe that directing consumers with questions about their genetic information to their health care professionals (a stance we continue to stand behind) is &#8220;misleading&#8221; because of a pronouncement by the Department of Health and Human Services&#8217; Secretary&#8217;s Advisory Committee on Genetics, Health and Society that physicians &#8220;cannot keep up with the pace of genetic tests and are not adequately prepared to use test information to treat patients appropriately.&#8221;
It is important to note that most instances of wrongdoing highlighted by GAO come from the 11 still unidentified companies they investigated. (Companies 2-4 were identified as DecodeMe, Pathway Genomics and Navigenics, respectively.)
The rest of the GAO report revolved around five main points (beginning on page 4 of the document):
(1) &#8220;each donor’s factual profile received disease risk predictions that varied across all four companies, indicating that identical DNA can yield contradictory results depending solely on the company it was sent to for analysis&#8221; 
It is well known that different DTC genetic testing companies provide different estimates of risk.   A report published in the scientific journal Nature in 2009 (<a href="https://www.nature.com/nature/journal/v461/n7265/full/461724a.html" target="_blank" rel="noopener">Ng, et al.</a>) described this situation, and 23andme, along with Navigenics, responded in an open letter that was <a href="https://blog.23andme.com/2009/11/18/23andme-navigenics-open-letter-to-nature/" target="_blank" rel="noopener">published on our blog, The Blog.</a>
There are valid scientific reasons for different estimates from different companies, such as: companies employ slightly different statistical models for making risk estimates; companies establish different criteria for the inclusion of associations in their reports; new associations are being discovered at a faster rate than companies’ development cycles; companies may test for an imperfectly overlapping set of genetic variants for reasons including the ability of different genotyping technologies to assay certain variants.
We agree with the idea that consumers should be able to compare the risk predictions they might receive from different test providers. This is an issue that deserves serious attention and one that we believe can be solved by the implementation of broad standards throughout the industry.   We have approached both the NIH and FDA for assistance in this matter (see this <a href="https://blog.23andme.com/2010/07/06/23andme-letter-to-heads-of-fda-and-nih/" target="_blank" rel="noopener">letter sent to the heads of both agencies and posted on our blog, The Blog</a>).   Instead of constructively adding to these efforts, GAO has instead implied that because results differ between companies, they are simply wrong.   Their report fails to provide all relevant information, and perpetuates the misunderstandings of genetics in particular and science in general that 23andMe has since the very beginning been dedicated to changing.
It should be noted that the problem of different risk predictions from different sources is not unique to the direct-to-consumer genetic testing industry.   Take for example, these two tools (<a href="http://www.reynoldsriskscore.org/" target="_blank" rel="noopener">here</a> and <a href="http://hp2010.nhlbihin.net/atpiii/calculator.asp" target="_blank" rel="noopener">here</a>) available to the public to calculate risk for cardiovascular disease.   Both are clinically valid, yet they give different answers.   It should be noted that development of both of these two tools was at least partially funded by the federal government.
2) &#8220;these risk predictions often conflicted with the donors’ factual illnesses and family medical histories&#8221;
As an example of this the GAO points to one of their &#8220;donors&#8221; who has a family history of heart disease but is told by some DTC genetic testing companies that he has average risk for heart disease based on genetics alone.
As the average risk of heart disease is about 21% for males, it is unsurprising that an individual would have a family history of heart disease but also be at average risk based on genetics alone.
The GAO admits that risks are probabilistic by definition and that therefore there is no expectation that there would always be complete alignment between predicted risk and reality.   But they then go on to quote an expert that &#8220;noted that the discrepancies between actual health and the predications made by these companies also serve to illustrate the lack of robustness of such predictive tests.&#8221;   This is a criticism that could be said of any risk prediction, including those currently made based on measurements such as cholesterol.   It is well accepted that not everyone with high cholesterol has a heart attack, and not all heart attacks happen in people with high cholesterol. Yet, cholesterol measurements are taken and used in clinical practice every day.
3) &#8220;none of the companies could provide the donors who submitted fictitious African American and Asian profiles with complete test results for their ethnicity but did not explicitly disclose this limitation prior to purchase&#8221;
Unfortunately, because of where and how biomedical research is funded, there are comparatively few studies looking for genetic associations in populations that form a minority in the countries where much of the latest research takes place. In other words, most of the associations have only been confirmed in populations of European ancestry.   An FAQ on our website explains this situation and lays out how it can result in fewer results for people with African American or Asian ethnicity.   In addition, our Terms of Service explain the limitations of genetic research, especially with regard to ethnic minorities, and note that a free demo version of our service is available for people to examine exactly what they will be getting if they choose to purchase.
23andMe actively looks for associations that have been repeated in populations other than the original study population, but we can only report data published by the scientific community at large. Unfortunately, most studies, both publicly and privately funded, do not include sufficient numbers of people from minority populations to meet the strict standards we have put in place to ensure that the information we provide through our Personal Genome Service is of high quality.
One of the strengths of 23andMe is our research platform, which we believe will be instrumental in bringing more diversity to the populations that can be completely served by DTC genetic testing.   We are actively seeking to initiate a large study of African American individuals that will seek to replicate genetic associations already known in European populations (or show that these associations do not hold).   As our database grows, we will be able to conduct studies in even more populations.
It should also be noted that the inability to accurately provide information for racial minorities is not a problem unique to the DTC genetic testing industry.   The <a href="http://www.cancer.gov/bcrisktool/" target="_blank" rel="noopener">Gail breast cancer risk estimator provided by the National Cancer Institute</a> has the following disclaimer:
&#8220;Assessments for Hispanic women are subject to greater uncertainty than those for white and African American women. Researchers are conducting additional studies, including studies with minority populations, to gather more data and to increase the accuracy of the tool for women in these populations.&#8221;
4) &#8220;one company provided donors with reports that showed conflicting predictions for the same DNA and profile, but did not explain how to interpret these different results&#8221;
This criticism specifically targets 23andMe and has to do with the fact that we provide both Established Research Reports and Preliminary Research Reports.   The distinction between the two is that Established reports are supported by multiple, large, peer-reviewed studies, whereas markers included in Preliminary reports have not yet gained enough scientific consensus to be included in the Established reports.
Very often there will be some markers for a particular disease that have reached &#8220;Established&#8221; status, while some have not.   In these cases there can be two reports (one Established, and one Preliminary) for the same condition.   This allows us to provide customers with a fuller picture of the scientific research, while at the same time indicating differing levels of confidence in different markers. 
The apparently conflicting predictions of an Established report and a Preliminary report are not inconsistent.   Consider smoking, weight and your risk of heart disease. Your weight might suggest a low risk for heart disease, while your smoking habits may suggest otherwise.   This example is analogous to cases of conflicting predictions in the GAO report, which rely on different genetic markers: one genetic variation may suggest an increased risk while another may suggest otherwise.
The GAO report does not reflect the great pains we have made on our website to make clear the differences between Established and Preliminary Research Reports.   We are, however, always looking for ways to improve our service, and we will take the feedback from GAO on this matter into consideration and look for even more ways to increase customer understanding.
5) &#8220;follow-up consultations offered by three of the companies provided only general information and not the expert advice the companies promised to provide&#8221;
This criticism does not apply to 23andMe because we do not have in-house genetic counselors.   We therefore can&#8217;t comment on the quality of the information provided by the &#8220;experts&#8221; the GAO had consultations with.
23andMe has always felt strongly that customers with questions about their data should consult with independent health care professionals so as to receive unbiased advice.   We provide a list of resources in every report directing customers to the National Association of Genetic Counselors, which maintains a website that can help people find local health professionals trained in genetics.
23andMe is also collaborating with Informed Medical Decisions, Inc., to give customers direct access to board-certified genetic counselors that have been specifically trained to guide them through their 23andMe results.   By design these counselors are independent of 23andMe, giving customers the opportunity to get unbiased information from a trained medical professional.
In conclusion, 23andMe is extremely disappointed that we did not have the opportunity to address all of these concerns at the Congressional hearing, or sooner, due to our lack of access to the report.   These are serious issues that deserve serious and thoughtful discussion.   Standards are needed in the genetic testing industry.   We have been working towards these since the inception of our company, and we were pleased to hear FDA say that they are interested in developing a new type of regulatory framework that can deal with the many special aspects of direct-to-consumer genetic testing while still providing consumers with the protections they need and deserve.   23andMe is meeting with the FDA today and looks forward to fruitful discussion.

GAO Studies Science Non-Scientifically

<img loading="lazy" decoding="async" class="alignright size-full wp-image-6586" title="23andMe_Logo_blog" src="https://blog-api.23andme.com/wp-content/uploads/2010/05/23andMe_Logo_blog.jpg" alt="" width="325" height="324" srcset="https://blogcms.23andme.com/wp-content/uploads/2010/05/23andMe_Logo_blog.jpg 350w, https://blogcms.23andme.com/wp-content/uploads/2010/05/23andMe_Logo_blog-150x150.jpg 150w, https://blogcms.23andme.com/wp-content/uploads/2010/05/23andMe_Logo_blog-300x300.jpg 300w, https://blogcms.23andme.com/wp-content/uploads/2010/05/23andMe_Logo_blog-271x271.jpg 271w" sizes="auto, (max-width: 325px) 100vw, 325px" />
Earlier today science writer Ed Yong wrote about his experience with 23andMe over at his blog, Not Exactly Rocket Science.   In this piece, he said that his 23andMe results showed that if he procreated with another science writer named Mark Henderson (a joke, as both are male), their child would not have cystic fibrosis (CF).
Several commenters were quick to point out that this is not true, that in fact Yong&#8217;s (and Henderson&#8217;s) 23andMe results merely show that they do not carry some CF mutations, not that they carry no CF mutations.
This is an important point and one that 23andMe stresses several times in our <a href="https://www.23andme.com/health/Cystic-Fibrosis/" target="_blank" rel="noopener">CF report</a>.   For almost every disease inherited in a simple recessive manner (meaning that a person must inherit a mutation in a certain gene from both parents in order to actually have the disease), there are many, many known mutations.   Some of these are more frequent than others, and the chance of having certain mutations often varies by ethnicity.
In a world where everyone has his or her full genome sequenced, it will be pretty easy to look for every known mutation associated with a certain disease.   But until that day, we&#8217;re left with imperfect tests that look at a selection of common mutations in order to screen for whether a couple might be at risk of having an affected child.   These tests can tell you that there is a chance that a child will be affected.   But if no mutations are detected, you don&#8217;t know that a child absolutely won&#8217;t be affected.   The presence of undetected mutations in each parent could still lead to a child being born with the disease.   Absence of evidence is not evidence of absence.
In the case of CF, more than 1000 mutations have been documented in the CFTR gene, and more than 500 of these have been shown to be disease-causing.   The American College of Medical Genetics recommends prenatal CF testing with a panel of the 23 most common mutations.   23andMe tests for these same 23 mutations, as well as eight others.   This information is available in our <a href="https://www.23andme.com/health/Cystic-Fibrosis/" target="_blank" rel="noopener">CF report</a>.
As a genetic counselor who contacted Yong pointed out in an email (that Yong then posted in the comments on Not Exactly Rocket Science), the mutations 23andMe currently tests for offer more coverage to people with European ancestry than they do for people with other backgrounds. We explain this in the CF report on our site:
<blockquote>Ancestry also affects the chances that a person carries a mutation not included in the 23andMe panel. For Jewish and non-Jewish Caucasians, this panel will identify 85-97% of all known mutations. Numbers are much lower for African Americans, Hispanics and Asians (50-75%). Bear in mind that the panel of mutations reported here is not exhaustive for any group and that if you are concerned about cystic fibrosis you should consult a medical professional.</blockquote>
We are constantly striving to make our reports more useful to our customers and to make important points more apparent.   We appreciate Yong&#8217;s efforts and those of the commentators on his blog to set the record straight.
23andMe is committed to giving people access to their DNA.   We&#8217;re also committed to giving people the tools they need to understand this information.   This incident highlights how important educating the public about genetics is.   As our General Counsel Ashley Gould pointed out in her comments to the <a href="https://blog.23andme.com/2010/07/20/23andme-at-fdacdrh-public-meeting-on-oversight-of-laboratory-developed-tests-ldts/" target="_blank" rel="noopener">FDA/CDRH public meeting on laboratory developed tests yesterday</a>, we believe that our service can play an important part in educational efforts, as learning in the context of your own genetic data can be extremely powerful.

Absence of Evidence Is Not Evidence of Absence: Understanding What Genetic Testing Can (And Can’t) Tell You

<img loading="lazy" decoding="async" class="alignright size-full wp-image-6586" title="23andMe_Logo_blog" src="https://blog-api.23andme.com/wp-content/uploads/2010/05/23andMe_Logo_blog.jpg" alt="" width="237" height="236" srcset="https://blogcms.23andme.com/wp-content/uploads/2010/05/23andMe_Logo_blog.jpg 350w, https://blogcms.23andme.com/wp-content/uploads/2010/05/23andMe_Logo_blog-150x150.jpg 150w, https://blogcms.23andme.com/wp-content/uploads/2010/05/23andMe_Logo_blog-300x300.jpg 300w, https://blogcms.23andme.com/wp-content/uploads/2010/05/23andMe_Logo_blog-271x271.jpg 271w" sizes="auto, (max-width: 237px) 100vw, 237px" />
As we have previously detailed here in the Blog, 23andMe received a letter from the Committee on Energy and Commerce of the United States House of Representatives asking for details of our genetic testing services.   We received a second letter from the Committee asking for more information about the incorrect processing of a number of customer samples that we previously discussed in the Blog on June 8, 2010.
If you would like to read the letters 23andMe received from the Committee, they can be accessed as PDFs through the following links:
<ul>
<li>Letter from the Committee on Energy and Commerce of the  United States House of Representatives–May 19, 2010</li>
<li>Letter from the Committee on Energy and Commerce of the  United States House of Representatives–June, 14, 2010</li>
</ul>
23andMe is working to provide Congress with the information they have requested.   Reprinted below is the letter we sent to the Committee on June 28, 2010 in response to their inquiry about the incorrectly processed samples.

June 28, 2010

Hon. Henry A. Waxman, Chairman 
House Committee on Energy and Commerce 
2125 Rayburn House Office Building 
Washington, DC 20515-6115

Hon. Bart Stupak, Chairman 
Subcommittee on Oversight and Investigations 
2125 Rayburn House Office Building 
Washington, DC 20515-6115

Hon. Joe Barton, Ranking Member 
House Committee on Energy and Commerce 
2125 Rayburn House Office Building 
Washington, DC 20515-6115

Hon. Michael C. Burgess, Ranking Member 
Subcommittee on Oversight and Investigations 
2125 Rayburn House Office Building 
Washington, DC 20515-6115


Re: June 14, 2010 letter request to Anne Wojcicki, President, 23andMe, Inc.

Dear Chairman Waxman, Ranking Member Barton, Chairman Stupak and Ranking Member Burgess:

On June 14, 2010, you sent a letter request to Anne Wojcicki, President of 23andMe, Inc. (“23andMe” or “the Company”), seeking information related to a June 8, 2010 announcement by the Company relating to a number of customer DNA samples.   While we are in the process of producing responsive documents, the Company would like to take this opportunity to explain the incident.   At the outset, we understand that a representative of the laboratory processing and analysis company that 23andMe uses will meet with you this week and will certainly be in the best position to discuss the issues that occurred within their facility.   Our description of events at the laboratory is limited to 23andMe’s own understanding of what happened there.

I.   How 23andMe’s Collection of Customer DNA Samples and NGI’s Processing and Analysis Operate

Preliminarily, we thought it would help to provide a step-by-step explanation of how 23andMe operates with respect to the collection of customer DNA samples and their processing and analysis by the Company’s third-party laboratory vendor, the National Genetics Institute (“NGI”) a wholly owned subsidiary of the Laboratory Corporation of America.

First, customers visit the 23andMe website and purchase one of three products for personalized genetic testing.   23andMe then sends to each customer a kit to collect a DNA sample, including a vial for saliva marked with a unique barcode, from which customer DNA will be extracted and analyzed by NGI.   The customer deposits a sample of saliva in the vial and then sends it directly to NGI.   The Company uses the unique barcode on each saliva collection vial to track and distinguish each customer’s unique data while maintaining customer privacy.

NGI is a Clinical Laboratory Improvement Amendments Program-certified laboratory that

provides advanced clinical genetics testing services for blood screening, medical testing, and clinical research.   Licensed as a clinical laboratory provider by both state and federal agencies, NGI participates in a number of approved quality control programs, and holds active Biologics Licenses from the U.S. Food and Drug Administration for screening of plasma for blood borne infectious agents.   It also provides advanced genetic testing services to physicians, hospitals, and clinics and has supported numerous pharmaceutical and biotechnology companies in the clinical development of new infectious disease and oncology therapies.   (see http://www.ngi.com/services/index.asp).   We understand NGI complies with strict professional, regulatory, and corporate quality-assurance standards.   23andMe employees and consultants do not participate in the processing and analysis of data at NGI, although we rely on the data returned.


Upon receipt of the saliva samples, NGI logs them by their assigned bar codes into NGI’s system, and then processes them through “amplification,” extracting the customers’ DNA from their saliva samples.   Each customer’s DNA is then placed by pipette into an assigned slot contained on a well plate.   The well plates used by NGI for testing of 23andMe customer samples are 5”x7”, and contain 96 slots to allow for the processing of up to 96 customer DNA samples per plate.   These well plates are then inserted into Illumina-platform equipment for analysis of nearly 600,000 data points that 23andMe provides to its customers.

After NGI completes the analysis, the collected data for each sample is linked to its accompanying bar code, and NGI electronically sends the data encrypted to 23andMe through a secure FTP site.   A single electronic file contains the data from one well plate.   Upon receiving this data, 23andMe matches each barcode to the customer it belongs to, and then uploads the data from each barcode to the customer’s individual account.   Finally, 23andMe sends emails to its customers notifying them that their data are loaded and ready for viewing.

II.   Factual Background of the June 2010 Incident

Summary

In brief, the June 2010 incident was caused by the human error of one of NGI’s certified technicians accidentally and wrongly inverting one well plate, which contained 96 customer DNA samples, nine of which failed processing by 180 degrees.   This led to the Company’s receipt of incorrect results, which were then transmitted to customers.   However, 23andMe responded rapidly to the problem, advised consumers of the problem, removed the inaccurate results in less than 24 hours, and provided those impacted with their correct results within six (6) days.   While 23andMe takes the incident seriously, the Company rapidly resolved it.   Since the incident, both 23andMe and NGI have further strengthened their quality control systems to avoid any further such problems.

Detailed Account 

The aforementioned inversion of the well plate by 180 degrees caused NGI’s computer to assign DNA information to incorrect barcodes, as the slots were assigned to barcodes in the computer before plate insertion.   I note that the plate inversion could have happened in any such lab for any such tests coming from any hospital, doctor’s office, or other NGI client — it was not a problem uniquely related to 23andMe samples.   Unfortunately, NGI did not detect the error before it transmitted the erroneous electronic data file from the inverted plate to 23andMe.   The Company received the NGI-generated, faulty data and uploaded it to each customer’s individual account and emailed those customers that their data was available.   Customers received these emails between 6 and 9 p.m. on the evening of June 1.

Early in the morning of June 2, 23andMe’s Director of Engineering, Alex Khomenko, checked his email and was alerted that several 23andMe customers had apparently received inaccurate genetic reports.   For example, in at least one instance, a customer had received genetic data indicating that the customer was the wrong sex.   That same morning, Mr. Khomenko contacted the Company’s Customer Service Manager, Michelle Klucsor, who said she had received similar customer complaints via email.

To identify what happened, Mr. Khomenko ran a program covering 100% of the samples received from NGI on June 1.   This program provided the names (which the customers provided when signing up with 23andMe) and sex (which the lab identified in its DNA analysis) of these customers.   He then compared a customer’s name against the lab-identified sex of the customer.   He noticed that, in some instances, the gender of a customer’s name (for example, “Sarah”) did not match the sex (for example, “male”) contained in the lab report.   He compiled a list of all such inconsistencies.   Mr. Khomenko then reviewed the original electronic files sent from NGI on June 1, 2010 and determined that all of the inconsistencies stemmed from a single electronic file, and, thus, he concluded, from one well plate.

As Mr. Khomenko conducted his investigation, 23andMe alerted NGI of the error.   NGI also conducted its own investigation.   Around 1 pm on June 2, 2010, 23andMe concluded that the data from the single file containing 87 customer data sets contained inaccuracies and began the process of removing this data from the accompanying accounts.   During this period, 23andMe continued discussions with NGI.   At approximately 4 pm on June 2, 2010 — less than 24 hours after customers had received the incorrect data &#8212; 23andMe deleted the erroneous data from the affected customers’ databases and sent emails to each affected customer alerting them of the error.

On June 5, NGI confirmed that its investigation had concluded that the issue stemmed from an error on a single well plate.   NGI’s investigation showed that the well plate had been inverted, causing the inaccurate customer data.   NGI had reprocessed the data of all impacted customers while its investigation was ongoing.   NGI re-ran the entire plate from both the amplified DNA samples and the original saliva samples provided by the customers.   In doing so, NGI was able to confirm the accuracy of the updated data.   23andMe was able to upload correct reports into the affected customers’ accounts on June 8, 2010.

III. Steps Taken by 23andMe and NGI to Ensure No Future Errors

Since this incident, NGI has informed 23andMe that NGI has taken steps to help ensure that such errors do not recur.   Specifically, 23andMe understands that NGI has modified the physical shape of its well plates to prevent them from being inverted in the machine that conducts the analysis.   Attached please find a photograph of the old well plate and one of the newly-designed well plates that are now in use.

For its part, 23andMe has implemented improved quality-control procedures to double-check the accuracy of NGI’s returned lab results before providing them to customers.   For the period of June 8 through June 23, 23andMe cross-referenced a customer’s name (which 23andMe predicts to be either male or female based on traditional name use) with the customer’s sex information pulled from lab data.   Doing so allowed 23andMe to identify potential errors, where the gender of a customer’s name is inconsistent with lab-identified sex data for that customer.

Starting June 24, 23andMe implemented a similar quality check with additional data.   As of June 10, 2010, 23andMe has been collecting customer date of birth and sex information pursuant to requirements that NGI is following.   23andMe’s new process cross-references this customer-provided sex data with that generated by the lab data.   Potential errors will be flagged where there are inconsistencies.

We look forward to speaking with you further about this issue.   If you have any questions in the meantime, please call me at (202) 339-8523.


Sincerely,



Michael J. Madigan 
Joshua P. Galper 
Stephanie Maya Cowles

Update on 23andMe and Committee on Energy and Commerce of the United States House of Representatives

[Update: This study was <a href="http://retractionwatch.wordpress.com/2011/07/21/sebastiani-group-retracts-genetics-of-aging-study-from-science/" target="_blank" rel="noopener">formally retracted</a> by <a href="http://www.sciencemag.org/content/333/6041/404.1.full">Science</a> on July 22, 2011]
A genome-wide association study of extreme longevity published last week in the journal <a href="https://dx.doi.org/10.1126/science.1190532" target="_blank" rel="noopener">Science</a> has been receiving a lot of press attention. &nbsp; The results are quite extraordinary: the authors identify 70 loci with genome-wide significant evidence for association with living past the age of 100, and they construct a SNP-based model for predicting exceptional longevity (EL) that has 77% accuracy in an independent set of individuals. &nbsp; We were initially very excited by the article and thought it would be of great interest to the 23andMe community. &nbsp; However after a closer reading of the article and supporting materials, we think this study actually inadvertently points to some of the pitfalls in analyses of genome wide datasets.
There are several reasons for skepticism about these new results. &nbsp; Another recent genome-wide study has reported no significant associations with longevity. &nbsp; There is suggestive evidence of genotyping quality control problems in the new results, and some routine quality control checks do not appear to have been done. &nbsp; The design of the study is particularly susceptible to introducing biases into the results. &nbsp; And a preliminary analysis of the proposed 150-SNP model for predicting longevity indicates that it is not predictive in the 23andMe community.
We expect that most of the results of this study will not have the same longevity as its participants. &nbsp; In genetics, as with most things in life, if a result seems too good to be true, it probably is. &nbsp; That said, this study does contain some interesting tidbits, such as the association of near the ApoE gene with longevity. &nbsp; This SNP has <a href="https://dx.doi.org/10.1371/journal.pone.0006501" target="_blank" rel="noopener">previously been shown</a> to be associated with Alzheimer&#8217;s Disease. &nbsp; For the time being we won&#8217;t be incorporating the data from this new study into our Personal Genome Service or putting information about any of the other SNPs here in the Blog. &nbsp; But we will be on the lookout for other attempts to replicate the study&#8217;s findings. &nbsp; If and when such a replication is published, we&#8217;ll scrutinize it like all of the papers we cover, and we’ll let you know what we find.
(A more technical discussion of the issues follows)
&#8211; A large study combining results of four genome-wide association studies of longevity was published in May in the <a href="https://dx.doi.org/10.1093/gerona/glq028" target="_blank" rel="noopener">Journals of Gerontology</a>. &nbsp; That study found no associations meeting their pre-specified criteria for genome-wide significance. &nbsp; While they used a more inclusive phenotype (age 90 or older), it is surprising that there could be so many loci associated with survival to age 100 in the new study, some with very large effect sizes, yet none were found in the larger study from earlier this year.
&#8211; An important part of designing an experiment is choosing the criteria that will constitute convincing evidence of a positive finding. &nbsp; We have to make a trade-off between setting the bar so high that we will miss many interesting and true results (false negatives), or setting the bar so low that we will get many spurious findings (false positives). &nbsp; In the new study, the authors use an unusually permissive standard for genome-wide significance that appears to allow a false positive rate of 6 per 100,000 SNPs tested, or 18 false positives across the whole study. &nbsp; A more conventional standard for significance is to control the false positive rate at 0.05, genome-wide, meaning that we expect only a 5% chance of even one spurious finding.
&#8211; None of the strong associations appear to be supported by any evidence from nearby SNPs. &nbsp; Each of the reported associations stands alone, but typically, we expect that nearby SNPs will show some intermediate evidence for association, because nearby SNPs tend to be correlated. &nbsp; This is a red flag because genotyping quality issues can produce these kinds of uncorrelated association signals.
&#8211; Many of the associated loci have high rates of missing genotypes in the EL individuals compared to the controls. &nbsp; For the 70 genome-wide-associated SNPs, the median missing data rate in EL samples was 9%, compared to 3% in controls. &nbsp; Some of the SNPs with high EL call rates appear to have large deviations from Hardy-Weinberg equilibrium in the EL group. &nbsp; For instance, the two SNPs with strongest evidence for association, and , are far out of equilibrium (P &lt; 10-20). &nbsp; Both of these are suggestive of data quality problems that can produce false associations, though there can be other valid explanations for the Hardy-Weinberg results.
&#8211; While the authors performed a replication of their results in an independent set of EL samples, the replication may share some biases with their initial genome scan. &nbsp; The authors drew most of their controls for their initial scan from a reference dataset, and used this same source of controls in their replication. &nbsp; And both sets of EL samples were genotyped with the same method. &nbsp; They show that substituting an alternate set of controls makes little difference, but this does not rule out a genotyping issue in the EL data.
We took a preliminary look in our customer data to see if the proposed SNP-based model described in Sebastiani et al. is predictive of exceptional longevity. &nbsp; A commonly used measure of test discrimination is to calculate how often, for a randomly selected case and control, a test correctly assigns a higher score to the case. &nbsp; This is known as the “c statistic” or “<a href="https://en.wikipedia.org/wiki/Receiver_operating_characteristic" target="_blank" rel="noopener">area under the curve</a>”. &nbsp; The authors of the new study say their model scored a 0.93 for this statistic. &nbsp; But when we compared 134 23andMe customers with age â‰¥ 95 to more than 50,000 controls, we obtained a test statistic of 0.532, with a 95% confidence interval from 0.485 to 0.579. &nbsp; Using 27 customers with age â‰¥ 100, we get a value of 0.540, with a 95% confidence interval from 0.434 to 0.645. &nbsp; A random predictor of longevity would give a 0.5 on this scale, so based on our data, performance of this model is not significantly better than random. &nbsp; Even with our small sample size, we can also clearly exclude values as high as the published result of 0.93.
Study designs that use independently collected control genotype data require extra attention to quality control to rule out the possibility of systematic differences in genotyping between cases and controls. &nbsp; In any experiment that tests hundreds of thousands of SNPs, a small proportion of SNPs can be expected to have problems with automated genotype assignment. &nbsp; This may not be a problem if cases and controls are affected equally, but if cases and controls are genotyped separately, then errors or missing genotypes may be concentrated in just the cases or just the controls, and these can give the appearance of a relationship between genotypes and phenotypes that is actually an experimental artifact.
The results could be strengthened if the authors could inspect the raw data for the associated SNPs to verify that genotypes are being assigned consistently. &nbsp;Differential missingness in the EL samples is a major issue because if it is the result of poor clustering, it will almost always tilt the apparent genotype frequencies of affected SNPs. &nbsp; If the clustering does suggest that there is a “batch effect”, then there are some strategies that can be used to rescue the analysis. &nbsp; One approach is to aggressively filter the data using strict quality criteria and manual inspection of putative associations. &nbsp; Another approach is to directly model the batch effect and then test for association conditional on the batch structure of the data. &nbsp; This only works if the problem is not perfectly confounded with the phenotype. &nbsp; If that is the case, the only resolution may be to use another technology to verify genotypes of associated SNPs and fill in missing values.
Addendum:
We repeated our analysis restricted to individuals with European ancestry. &nbsp; The results were similar: for 129 customers with age â‰¥ 95 and more than 43,000 controls, we got a test statistic of 0.534, and for 26 customers with age â‰¥ 100, we got a value of 0.558. &nbsp; In both cases, the 95% confidence interval includes 0.5.

Latest Posts — Page 173