DNA and You

If You Think Wired Is a Good Place to Go for Medical Advice, Think Again...

Andrew Yates at Think Gene wrote today about the new Wired Wiki home genomics how-to guide, "Check Yourself for Genomic Abnormalities."  Check out Andrew's post for a great discussion of the maturity, or, rather, the lack thereof, of the personal genomics market.

Something caught my eye though as I read through "Check Yourself for Genomic Abnormalities" at the Wired Wiki site.  The wiki post describes several options for "checking yourself for genomic abnormalities": 1) Visit a Genetic Counselor; 2) Scan Your Whole Genome; and 3) Perform Lab Tests at Home. 

Interestingly, the author(s), who otherwise did an ok job of briefly explaining what genetic counselors do, utilized consideration of a diagnosis of celiac sprue as an example of a situation in which someone would want to see a genetic counselor rather than "scanning their whole genome" or "performing lab tests at home."

I think the world of Wired, but in case it is not clear to the early adopters out there...

Wired is probably not where you want to get your medical advice.

Celiac disease (aka gluten-sensitive enteropathy or non-tropical sprue), the condition mentioned in the hypothetical scenario, is diagnosed via a blood antibody test and small intestinal biopsies.  Thus, rather than seeing your local genetic counselor if you think you might have CD, you would do well to discuss it with your primary care doctor and a gastroenterologist. 

The wiki writer's confusion likely stems from the fact that genetic factors do play a role in risk for Celiac disease; however, the genetics are complex, and the genes involved are not deterministic.  For example, risk of Celiac disease is higher if you have certain forms ("alleles") of HLA genes.  About 30% of the population has one of the Celiac disease-associated HLA alleles; however, only 3% of individuals with the Celiac disease-associated allele develop CD.

High Throughput Sequencing, Moore's Law, The Next Generation, and Storage Options

It will be interesting to see if DNA sequencing technology continues to follow a Moore's law-like trajectory over the coming decades.  Clearly, next generation technologies are going to have a massive impact both in the research setting and on medical resequencing. 

Alexis Madrigal at Wired Science wrote a nice piece on the technological advances leading to the current generation of high-throughput sequencers and also the next generation: Pacific Biosciences and Helicos.

Also notable is a post from the always excellent Daniel MacArthur at Genetic Future on options for storage of personal genome sequences.

Genetics Takes Over The New England Journal of Medicine

One only has to briefly scan the table of contents of tomorrow's issue (Jan. 10) of The New England Journal of Medicine to figure out that 2008 is going to be a big year at the crux of genetics and medicine!  The issue includes the following (note that only a subset of the following full articles are available without subscription):

• A perspective by Drs. David Hunter, Muin Khoury, and Jeffrey Drazen on the medical implications - or lack thereof - of personalized genotyping services (i.e., 23andMe, Navigenics, and deCodeMe).  More on this in a follow-up post later this evening.  However, I can tell you that these three are not fans of personalized genotyping companies.  There is also an audio interview with Dr. Khoury available here.
• Dr. John Bissler and colleagues from Cincinnati Children's Hospital Medical Center present the results of a study of sirolimus treatment of angiomyolipomas in tuberous sclerosis complex (TSC) and sporadic lymphangioleiomyomatosis.  TSC is a Mendelian genetic disease in which the genetic defects lead to constitutive activation of the "mammalian target of rapamycin" (mTOR, a key cellular signaling pathway intermediate).  As sirolimus suppresses signaling through mTOR, this study represents a rational use of sirolimus to treat angiomyolipomas in TSC.  More on this soon at my other blog, Cancer and Your Genes.
• Dr. Antonio Pelliccia and colleagues present the results of a study looking at implications of EKG abnormalities referred to as "repolarization abnormalities."  They show that out of 81 athletes with a particular type of EKG abnormality (see free full text here for details), 5 (6%) ultimately developed cardiomyopathies (including one individual who died from arrhythmogenic right ventricular cardiomyopathy - which has a genetic basis). 
• Dr. Melanie Percy and colleagues demonstrate that an oxygen sensing gene called HIF2A is mutated in a family with Familial Erythrocytosis (i.e., a heritable condition in which affected individuals have too many red blood cells).
• There is also a review of the book, "Reprogenetics: Law, Policy, and Ethical Issues," edited by Lori P. Knowles and Gregory E. Kaebnick.
• As if all that were not enough, an article in the NEJM "Clinical Problem" series focuses on the approach to Long QT syndrome, an inherited, genetically heterogeneous condition that predisposes individuals to life-threatening arrhthymias (abnormal heart rhythms). 
• Last, but certainly not least, in an online article published today, Mark Daly and colleagues report the identification of a small, sub-microscopic region of chromosome 16 that when deleted or duplicated leads to autism susceptibility!  Although this is probably only responsible for about 1% or so of cases, this is a huge accomplishment.

In looking at just this single issue of NEJM, I think it is safe to say that we have a very interesting year ahead of us.  Stay tuned to DNA and You for more detailed posts on the above!

KNOME, George Church, and Personalized Whole Genome Sequencing

I just received the following (copied and pasted) information in an email from Knome:

Human Whole-genome sequencing hits commercial market

20 individuals to be among first in history to be fully sequenced

CAMBRIDGE , Massachusetts —  Nov. 29, 2007  ­­­ — Knome, a personal genomics company, today announced the launch of the first commercial whole-genome sequencing and analysis service for individuals.

"In 2003, the Human Genome Project completed a 12-year effort to sequence the first human genome at a cost of $3 billion. Only very recently have costs come down to a level where it is now feasible for private individuals to be sequenced and analyzed. We expect this evolution to quickly usher in a new era in personalized medicine," said Dr. George Church, PhD, a co-founder of the firm and Professor of Genetics at Harvard Medical School.

First to know, first to benefit

Knome today opens enrollment for its first sequencing flight. Because the sequencing and analysis process is both labor and computationally intensive, initial capacity is expected to be limited to approximately 20 clients.

"To date, Craig Venter and James Watson are the only named individuals to have their genome sequenced. Our first 20 clients will have a historic opportunity to help pioneer the emerging field of personal genomics. They will be among the first to know and the first to benefit from the latest advances in our rapidly developing understanding of the human genome," said Jorge Conde, the firm's CEO.

Building the gold standard

Whole-genome sequencing decodes the 6 billion bits of information that make up an individual's genome. Unlike existing genome scanning or "SNP chip" technologies that provide useful but limited information on approximately 20 conditions, whole-genome sequencing allows for the analysis of up to 2,000 common and rare conditions, and over 20,000 genes – numbers that are rapidly growing.   

"Whole-genome sequencing is the endgame," according to Mr. Conde. "It will enable us to look at nearly 100% of your genetic code compared to the less than 0.02% currently available on SNP chips.  This is the approach that most fully reveals what our genomes can tell us about ourselves."

Pricing for Knome's service will start at $350,000, including whole-genome sequencing and a comprehensive analysis from a team of leading geneticists, clinicians and bioinformaticians. This team will also provide continued support and counseling. 

"Knome's goal is to establish the gold standard in personal genomic services for individuals. We are bringing our clients the latest sequencing technology, Knome's proprietary analytic engine and security solutions, and access to top genomic scientists and medical professionals," said Conde. "Analytics, privacy and on-going client service are as important to us as the actual sequencing."

Core to the fundamental principles of the company, clients will retain full ownership of their personal genome and have the ability to anonymously share all or portions of their genome with researchers and other medical professionals.

About Knome

Based in Cambridge, Massachusetts, Knome has the distinction of being the first personal genomics company to commercially offer whole-genome sequencing and analysis services for individuals. Working alongside leading geneticists, clinicians and bioinformaticians from Harvard and MIT, Knome enables its clients to obtain, understand, and share their genomic information in a manner that is both anonymous and secure. Knome is a privately funded company. Please visit www.knome.com for more information.

Knome is a trademark of Knome, Inc. All other company and product names may be trademarks of the respective companies with which they are associated.

For more information CONTACT

info@knome.com

Clearly, it will be interesting to see how this plays out.  I would be particularly interested to know what the specific plans are for the implied analysis and counseling from clinicians and other team members.

Anybody got a spare $350K they would like to part with?

NY Times Personalized Genomics Feedback

For those who haven't seen it yet, a series of letters responding to the NY Times coverage of 23andMe, Navigenics, and deCODE Genetics are available here.

Interestingly, the author of the last letter in the NY Times series, Dr. Hugh Rienhoff, has made some news himself of late. 

Who's your daddy?

Once a year, I teach a case-based discussion section in the medical genetics course for the 2nd year med students at the University of Washington.  Each year, there is an anomalous result in one of the cases that has several potential explanations.  We encourage the students to think through the possible explanations...and one of them is non-paternity.  That case sticks out in my mind, because one year, a student suggested that I was being sexist by suggesting that non-paternity was a possibility (i.e., if I bring up non-paternity...I must propose non-maternity as an option as well).  The class laughed and another student pointed out that generally maternity is a little more difficult to fake than paternity.  Now I look back on this with some amusement. 

Over the past decade, (non)paternity has really been in the public eye - via Jerry Springer and other daytime TV, via court room drama, and via the increasing availability of cheap paternity testing.  In my clinical genetics training, an oft quoted number was that ~5% of individuals in a clinic were not fathered by the person who is ostensibly their father.  I've never been able to track down the primary reference for this; however, estimates in the literature range from a minimum of ~1% to ~20%.  Regardless of the specific number, it is a substantial fraction of people.  The beginning of an era of personalized direct-to-consumer genomics that may be utilized by a substantial fraction of the population may transform the way we think about this issue.  At the very least, if companies like 23andMe and deCODE Genetics are successful in getting large numbers of families to utilize their genotyping services, there will be more...umm...transparency brought to this issue. 

I'm confident that these companies have spent a lot of time thinking about this issue, and I'm not suggesting that it is a reason to regulate or ban the personalized genomic testing by any means.  Nevertheless, inadvertent demonstration of nonpaternity is another way in which the personalized genomics revolution will affect the social fabric of our world.

What a week...

What a wild couple of days...

First, deCODE Genetics unstealths its deCODEme foray into the personalized genomics market.

Then, just when you think they've been completely beaten to the punch, 23andMe throws a counterpunch.

I'd been wondering for a while what deCODE was up to in this arena, as they clearly have the infrastructure and the track record to make a serious play in personalized genomics.

Interestingly, even though many would argue that the Navigenics approach is the most responsible one, getting results on 20 or so SNPs is looking so "last week" right now. 

Just in case you've been in a sound-proof bubble the last few days, here are some links to a few articles and posts about the events of the last few days:

• NY Times article about deCODEme launch
• CNN Money on deCODEme launch
• NY Times article about 23andMe launch
• Wired article mostly focusing on 23andMe
• Garett Rogers post

While much has been said about the product launches, I find it interesting that there has been relatively little talk about something that is being left out of the products offered to the public (at least at the moment as far as I can tell).  Specifically, that something is copy number variation. 

For those not familiar with the concept, in addition to variation at single DNA base pairs, it is also crystal clear that individuals differ enormously from each other in terms of the presence or absence or elevated copy number of entire large "chunks" of DNA.  These studies, pioneered by Evan Eichler and others, have shown that our genomes are more dynamic than was previously recognized.  Thus, when copy number variation involves important genes or other important stretches of DNA not in genes, it can have an important impact on disease susceptibility.  For example, copy number for a stretch of DNA containing the CCL3L1 gene has been beautifully shown to very significantly influence HIV/AIDS susceptibility.  Likewise, low copy number of a gene called FCGR3B is associated with the development of the kidney disease, glomerulonephritis, in individuals with systemic lupus erythematosus.

Although first generation SNP genotyping approaches didn't always do so well with detecting copy number variation, more recent iterations have the ability to detect copy number variation (although a different technique called array comparative genomic hybridization is more commonly utilized to detect this clinically).  For example, Affymetrix states on their website that their 500K arrays provide copy number variation information (Navigenics is partnering with Affymetrix for genotyping).  Likewise, Illumina (23andMe's genotyping partner) points out that their product can also identify copy number variants.

Although the launch of genotyping services by 23andMe and deCODE Genetics appears likely to provide dramatically more information to the consumer than the small number of SNP results that will be reported by Navigenics, it would appear that both 23andMe and deCODE may have access to information related to copy number variation that can have important implications with respect to disease risk and which may not be reported back to the consumer.  It will be interesting to see how this is dealt with in the future.

Eye Color and Your Genes

Although it was once thought to be a classic Mendelian trait, eye color is now well known to be subject to polygenic (i.e., multiple genes are involved) inheritance.  Our current catalog of genetic variation affecting eye color is incomplete. 

However, in a paper published online today in Nature Genetics (abstract here, subscription required for full paper), Dr. Kari Stefansson and colleagues at deCode Genetics report the results of a genome-wide association study focusing on eye color (in addition to hair color and freckling) as a phenotypic trait.  This study adds considerably to our knowledge of the genetics of human eye color, in addition to hair color and freckling. 

The authors initially studied about 3,000 Icelanders and then performed replication studies in another ~2700 Icelanders and about 1200 Dutch individuals.  With respect to eye color, they found the previously reported association with OCA2 gene variants - not big news but providing confidence in the methods.  However, the authors also found association of a new gene - SLC24A4 - with both eye and hair color.  Additionally, two coding variants in the TYR gene were found to be associated with eye color and freckles. 

Interestingly, the SLC24A4 gene is similar to SLC24A5; a SLC24A5 coding single nucleotide polymorphism has previously been shown to be associated with skin color in African-Americans and African-Caribbean populations. 

The authors then went on to derive a model in which they attempted to predict eye color based on genotypes.  They showed that variants in OCA2 dominate the distinction between brown and blue eyes and that the newly reported variants can contribute significantly to the more subtle distinction between blue and green eyes. 

It will be interesting to see what deCode may have in mind with respect to commercialization of this new information.

New Blog: Cancer and Your Genes

I've started another blog, "Cancer and Your Genes," which can be found here.  It will be focused on the intersection of our genomes and cancer risk, with particular emphasis on educating the lay public about genetic risks for cancer.

Mountains Out of Molehills? Or the Real Deal?

Despite some significant challenges in bringing effective personalized medicine approaches to the clinic, cancer treatment is one area in which personalized medicine and molecularly-targeted therapies have begun to come to fruition already.  In addition to Gleevec, which effectively targets the BCR-ABL fusion protein that is characteristic of chronic myelogenous leukemia (CML) in humans, Herceptin is important in the treatment of the subset of breast cancer patients whose cancers have substantial expression of the "Her2/Neu" protein.

However, these molecularly targeted therapies that focus on specific proteins important to the causation and maintenance of a subset of cancer cases are only available for a small percentage of all cancers.  To move closer to truly targeted therapy (which hopefully would have less severe side effects than current approaches with relatively non-specific cytotoxic drugs and radiation), we must know more about the relatively complicated changes in the DNA of the cancer cell.

Although it frequently seems like the more we learn about cancer the more complex the problem becomes, a paper published in the online edition of Science today makes a significant contribution to our knowledge of what goes awry in the development of breast and colon cancers.  (Although the full paper is available only by subscription, there is a press release with more information about the findings at the HHMI website) 

In this work, groups led by Giovanni Parmigiani, Kenneth Kinzler, Victor Velculescu, and Bert Vogelstein utilized a high-throughput DNA sequencing approach to look at the coding sequences (the DNA sequence coding for protein sequence) of more than 18,000 genes in each of 11 different breast cancer cases and 11 colon cancer cases.  Utilizing a careful approach to determine that mutations occurred somatically in the development of the cancer (as opposed to being polymorphisms present in all of the DNA of the individual at birth), the study basically showed that an impressively high percentage of the genes had a non-silent mutation (i.e., one that altered the protein coding sequence) in at least 1 cancer case: 9.4% of the total number of genes analyzed.  Putative cancer-related genes were evaluated in another set of cancers.  There were a total of 280 genes (equally distributed between breast and colon cancers) in all that were validated by the presence of mutations in both the initial cancer set and the validation tumor set.  Further analyses were performed to assess the plausibility that these genes were mutated more often in the cancers than would be predicted by chance. 

Although it is not news that cancer is a genetic disease associated with the accumulation of mutations in several key genes (that presumably differ by tumor/tissue type), the high-throughput resequencing approach utilized in this work suggests much more complexity in the mutated cancer genome than was previously recognized.  By plotting a score related to gene mutation frequency for each gene on something resembling a topographical map, the authors point out that while there are certainly still several "mountains" that appear to likely to be very important in cancer development, perhaps the more striking feature is an extremely high number of "hills" (genes mutated at a much lower frequency in a given cancer type, but which still likely play a role in causation or maintenance). 

Although some may suggest that the authors are making mountains out of molehills, I believe that their interpretation is likely correct.  The ability to sequence virtually the entire coding genome of a tumor is very exciting; with the coming drop in sequencing costs, this could be more realistically applied to individual patient tumors in a clinical context.  The hard part will be devising strategies to act on this information in a way that improves clinical outcomes.  However, as the authors suggest, many of the "hills" can be grouped into one of a small number of biochemical pathways.  Future studies will no doubt assess whether drugs targeting these pathways can be applied on a rational basis based on high-throughput sequencing of patient tumor DNA. 

Despite some significant challenges in bringing effective personalized medicine approaches to the clinic, cancer treatment is one area in which personalized medicine and molecularly-targeted therapies have begun to come to fruition already.  In addition to Gleevec, which effectively targets the BCR-ABL fusion protein that is characteristic of chronic myelogenous leukemia (CML) in humans, Herceptin is important in the treatment of the subset of breast cancer patients whose cancers have substantial expression of the "Her2/Neu" protein.

However, these molecularly targeted therapies that focus on specific proteins important to the causation and maintenance of a subset of cancer cases are only available for a small percentage of all cancers.  To move closer to truly targeted therapy (which hopefully would have less severe side effects than current approaches with relatively non-specific cytotoxic drugs and radiation), we must know more about the relatively complicated changes in the DNA of the cancer cell.

Although it frequently seems like the more we learn about cancer the more complex the problem becomes, a paper published in the online edition of Science today makes a significant contribution to our knowledge of what goes awry in the development of breast and colon cancers.  (Although the full paper is available only by subscription, there is a press release with more information about the findings at the HHMI website) 

In this work, groups led by Giovanni Parmigiani, Kenneth Kinzler, Victor Velculescu, and Bert Vogelstein utilized a high-throughput DNA sequencing approach to look at the coding sequences (the DNA sequence coding for protein sequence) of more than 18,000 genes in each of 11 different breast cancer cases and 11 colon cancer cases.  Utilizing a careful approach to determine that mutations occurred somatically in the development of the cancer (as opposed to being polymorphisms present in all of the DNA of the individual at birth), the study basically showed that an impressively high percentage of the genes had a non-silent mutation (i.e., one that altered the protein coding sequence) in at least 1 cancer case: 9.4% of the total number of genes analyzed.  Putative cancer-related genes were evaluated in another set of cancers.  There were a total of 280 genes (equally distributed between breast and colon cancers) in all that were validated by the presence of mutations in both the initial cancer set and the validation tumor set.  Further analyses were performed to assess the plausibility that these genes were mutated more often in the cancers than would be predicted by chance. 

Although it is not news that cancer is a genetic disease associated with the accumulation of mutations in several key genes (that presumably differ by tumor/tissue type), the high-throughput resequencing approach utilized in this work suggests much more complexity in the mutated cancer genome than was previously recognized.  By plotting a score related to gene mutation frequency for each gene on something resembling a topographical map, the authors point out that while there are certainly still several "mountains" that appear to likely to be very important in cancer development, perhaps the more striking feature is an extremely high number of "hills" (genes mutated at a much lower frequency in a given cancer type, but which still likely play a role in causation or maintenance). 

Although some may suggest that the authors are making mountains out of molehills, I believe that their interpretation is likely correct.  The ability to sequence virtually the entire coding genome of a tumor is very exciting; with the coming drop in sequencing costs, this could be more realistically applied to individual patient tumors in a clinical context.  The hard part will be devising strategies to act on this information in a way that improves clinical outcomes.  However, as the authors suggest, many of the "hills" can be grouped into one of a small number of biochemical pathways.  Future studies will no doubt assess whether drugs targeting these pathways can be applied on a rational basis based on high-throughput sequencing of patient tumor DNA.