Here’s how it goes for me: a few afternoons a year, usually when I haven’t slept or eaten right, but sometimes for no apparent reason, I begin to sense a pressure behind my left eyebrow and to feel queasy. By now I know what’s coming, and I resign myself to another miserable evening and a coming day or two lost to indistinctness. I rush home and secrete myself in the coolest, darkest spot I can find, because for each of my senses the volume seems to have been cranked to amphitheater-level. I lie there for four or five hours, a dog on a leash, thinking grim thoughts and, despite myself, yelping every now and again when the pain ratchets up. Perhaps you know somebody with migraine and are familiar with the vocabulary they use to capture the experience: ‘throbbing’, ‘nauseating’, ‘excruciating’ and the like. All true. Respite comes only when my stomach has had too much and returns my lunch — normally one wants to avoid this outcome, but here I welcome it, court it even, which I’ve always found darkly funny. Then I fall into a dreamless sleep. While some don’t have it as bad as me, many have it far worse.

With the launch of our new migraine headache survey today, we at 23andMe invite you all to share your headache experiences, whether you’re one of the lucky few who’s never had even a little one or someone who must deal with the threat of migraine pain on a daily basis.  You needn’t be a 23andMe customer to take the survey (although we recommend it). All you need is a free 23andMe account.

Migraine headaches are nasty things. The common feature is a terrible pulsing pain emanating from inside the skull, usually just on one side, but apart from this everyone experiences them a bit differently. Some unlucky folks get them every day, while others get them just once a year. Migraines can last for a few hours or can pound on for days at a time. Then there is the menagerie of symptoms that can accompany the headaches, including nausea, vomiting, visual or aural illusions, and aversion to light, smell, touch and/or sound. Perhaps most variable across people are the causes of the headache, or triggers. For one person the triggers might be red wine or nuts, for another they might be stress, bright lights, or noise.

There is a wide array of treatment options for migraine. With guidance from their doctors, most migraine sufferers nowadays are able to find partial or full relief from their headaches. Despite the effectiveness of these treatments, the basic biology of the disease is not well-understood¹,  and migraine continues to exact a tremendous physical and economic toll on our society².

Two prominent migraine researchers have suggested that the blame for the slow progress in understanding migraine lies with a systemic lack of public funding for migraine research. They argue that the relatively recent, and incomplete, acceptance of migraine by the medical and research communities as a genuine medical problem, as opposed to mere melodrama, has led migraine’s funding to lag well behind that for diseases of similar impact. For example, they estimate that while $13.80 is spent for each sufferer of asthma, just 36 cents of federal research funds are spent per migraine sufferer.

The genetics of migraine are also only partially understood. That’s where our new survey comes in. Our community-based research program 23andWe seeks to empower the public to engage in genetic research from the ground up. We know our efforts cannot substitute for proper federal support of migraine research, but evidence of great public interest, plus a new finding or two, would add to our understanding of the disease and potentially send a message to Washington.

With all haste, then, please head over to the new migraine survey and be counted!

Footnotes:

1. What is understood of its biology and chemistry is fascinating, and summarized well here.
2. Nearly 40 million people in the US, and a similar number in Europe, suffer from migraine, roughly one in every ten people. Migraine occurs in women about three times more commonly than in men. Migraine is estimated to cost  around $23BN/year in the US and Euro27BN/year in Europe in direct medical costs and in indirect costs, such as lost productivity.

ShareThis

Nearly a year ago today, the Spittoon reported that researchers had figured out how to pick a known person’s DNA out of a database containing genetic information for up to 200 unidentified individuals. The feat made it possible to determine, for example, if a particular suspect’s DNA was present in a mix of several peoples’ collected at a crime scene. It also raised the possibility, at least theoretically, that somebody could figure out if a particular person’s DNA was part of a research database or some other pool of anonymous data.

Now another team of researchers has calculated the limits of this needle-in-a-haystack technique. It turns out that if the number of data points revealed is small enough, an individual’s DNA can remain undetectable.

Before UCLA geneticist Stan Nelson and his colleagues published their paper last year, it had been a common practice to publish summary data from research DNA collections so members of the genetics community could build on each others’ research. (Note: 23andMe does not share individual-level data without explicit consent, and is working on ways of distributing summary data without compromising privacy.) It was assumed that aggregation would make it impossible to tell whether a particular individual had contributed to that collection. But in the wake of the announcement of the new detection method, major genome centers swiftly removed summary data from public view to avoid the possibility of compromising the identity of any research participants.

In a paper published today in Nature Genetics, researchers from UC Berkeley take the first step towards a less extreme route to protecting individuals’ identities than simply withholding all information. The researchers developed methods to calculate precisely how much genetic information from a research DNA collection may be revealed without risk of exposing the identities of the study participants.

At the heart of the PLoS Genetics paper that started the hubbub was a statistical test that the authors used to show that they could reliably tell whether someone was present in a DNA mixture. The new study not only provides an improved version of this test, but proves that the new approach is the best of all possible detection methods. Since no better test could be devised, if theirs is unable to detect whether an individual was part of a mixture of DNA, they can be confident that no test would be able to do it.

Armed with this certainty, the authors explore just how much information from a research DNA collection may be reported without risk of divulging who was in the study. It turns out that for a collection of 1,000 individuals, up to 10,000 individual data points can be revealed with little risk of exposing anyone to identification.

This makes it somewhat easier for geneticists to share their data. But these days an individual’s genetic data typically consists of a genome-wide panel of 500,000 to a million single-letter DNA variations known as SNPs — so the problem still isn’t solved completely.

ShareThis

Malaria is one of the leading causes of death in the developing world, claiming nearly a million victims each year. The great majority of them are African children below the age of five. The illness is caused by a single-celled parasite called Plasmodium that is transmitted by mosquito bites to humans. In a paper published today in Nature Genetics, a group of African and British doctors and scientists report on their study of the genetic roots of malaria susceptibility. They found no new smoking gun with this effort, but learned much about how to improve African genetic studies in the future.

The researchers gathered the SNP genotypes of 2,500 children, with the consent of their parents, from a small region in The Gambia. About 1,000 of the children had been admitted to the hospital with a case of severe malaria — the other 1,500 were newborns. In a genomewide association study, the researchers checked each of a half-million SNPs (single nucleotide polymorphisms) for sharp differences in genetic composition between the group of children suffering from malaria and the group of newborns, who served as an approximation of a malaria-free group. If one version of an individual SNP was seen at high frequency among the malaria victims, but at low frequency in the newborns, then the difference might be because the SNP tends to cause malaria or is nearby one that does.

Upon scanning their data, the researchers came up more or less empty-handed: by the usual standards of the field, none of the 500,000 SNPs would pass muster.

This deflating result stands at odds with what is known already about the genetics of malaria susceptibility. Most people who have taken a biology class learn that human populations in malarial regions have developed a natural immunity to malaria infection, not through their immune systems, but through a genetic modification of hemoglobin. Hemoglobin is a molecule charged with ferrying oxygen from your lungs (and the lungs of most life forms that have them) to all your cells, an essential task. Biologists have traced hemoglobin-based malaria resistance to a change at a single DNA base pair on chromosome 11 — wouldn’t we expect at least this SNP to light up as significant?

In truth, the failure wasn’t so surprising; it arises from the interplay of genetics with our species’ history. Humans first arose in Africa, so that’s where genetic variation has had the longest time to build up. Modern-day Asian, European, and Native American people descend from people who emigrated from Africa about 50,000 years ago. These migrants carried just a subset of the African gene pool with them, so non-African populations today have much less “well-mixed” genomes than African populations. The present study uses genotyping chips developed for use in European populations, and its failure to find the known hemoglobin SNP (which isn’t even genotyped by the chip) and other known genetic contributors to malaria resistance is essentially due to the fact that you’d need more like two million SNPs than half a million to do the job right.

The solution, you’d think, is just to make a chip with a lot of markers for specific use in Africa, and be done with it. But the authors show that African genomes appear to be mixed so well that no single such chip could be designed. Instead, they propose an alternative approach: use a good but inevitably suboptimal African SNP chip in your full study sample, then obtain full genome sequences from a small number of the members of that sample. Then, using a powerful statistical method called imputation, you use the full sequences of the smaller group to fill in the full genomes of the entire study sample based on their SNP genotypes. This approach, as the authors demonstrate convincingly in the case of hemoglobin-based malaria resistance, would provide a statistically powerful and economically viable means of tracking down the causes of some of the most challenging health problems of our time.

ShareThis

Map of Mexico with sampled-from states highlighted.

Recent studies of genetic diversity among Europeans (blogged about here and here) show that DNA is a surprisingly good predictor of where a person lives; people from the same country tend to be more similar to one another than to those from other parts of the continent. This latest study, which was published earlier this week in the Proceedings of the National Academy of Sciences, shows a similar pattern in Mexico: Mexican Mestizos (people of mixed European and Native American ancestry) from the same state tend to group together genetically, and the groups themselves fall along a genetic continuum that corresponds roughly to their latitude. You can see for yourself in this plot from the paper (below), which we’ve modified a bit for clarity. This is the same kind of plot used in 23andMe’s Global Similarity: Advanced feature — each point in the plot represents a person. The closer two points appear in the plot, the closer those two individuals are to each other genetically. The 300 Mexican Mestizos fall into a line stretching from a group of Europeans at the upper right to a group of Amerindians¹ at the lower left.

PCA map of Mexican genotypes.

The people from Sonora, the northernmost state, appear closest to the European cluster, and the people from the sourthern states Guerrero, Veracruz, and Yucatan appear closest to the Amerind cluster. This makes you wonder whether this pattern might correlate with the proportion of European ancestry. The researchers wondered that too, so they investigated further by analyzing their dataset with a computer program that estimates the proportion of ancestry a person’s DNA derives from each of several reference populations. When they set the program loose, they found that the six states did vary widely in proportion of European ancestry, from an average of 65% in Sonora (fifth column from the left) to an average of 35% in Guerrero (second from the right):

Admixture proportion estimates for Mexican and HapMap samples.

The authors note that this pattern makes sense, since Amerindian population density declines as you head north. Also, you might note that there’s a green sliver of African ancestry in each of the Mestizo populations, which approaches 5% in the southern states of Veracruz and Guerrero. This also meshes well with the historical record, since it’s through these coastal states, among others, that African slaves entered Mexico during the Spanish colonization.

These analyses tell us much about Mexico’s history — could the same dataset serve to improve Mexico’s future? This paper also examines the prospects for conducting medical genetic studies in Mexico.

The most popular study design nowadays, and the one that the researchers focus on, is the genomewide association study, or GWAS, also called a case-control study. GWASes are the basis for essentially all the studies discussed in the Spittoon’s SNPWatch section, and the majority of the findings underlying 23andMe’s Health and Traits reports, and have recently been the subject of intense debate in the genetics community.

The core idea of a GWAS is to look for genetic markers (these days, usually common single-letter DNA variations known as SNPs) in a population that are at very different frequencies in a group of people with a particular disease, say type 2 diabetes, than in a group of people without that disease. In order to do that, you first have to identify a comprehensive set of the common genetic variations within a population, and then create a DNA array (commonly called a SNP chip) to probe all those variable locations in a large number of people with and without the condition you’re studying.

There’s the rub. In recent years, a project known as HapMap has created catalogs of common variations in European, African and East Asian populations, and chips have been produced based on it². But Mexico’s population is a mixture of two of those (European and African) and another population (Amerindian) that is related, but not identical to, the third. There is no Mexican SNP chip.

That’s why the authors of this paper are suggesting a project to characterize common genetic variation within the Mexican population itself. They estimated that a catalog of common genetic variation using any two of the Mestizo groups they analyzed would capture enough variation to fuel quality GWAS studies, and would require fewer markers to do so than the alternative, which would be to use all the common markers from HapMap itself. This would substantially lower the cost of genotyping, they argue, and the reduced cost of using a platform based on the Mexico-specific catalog would allow researchers to genotype many more people for their GWAS studies for the same number of pesos. Since sample size is often the limiting factor in the ability of the GWAS design to find disease genes, this could improve their ability to find the genetic causes of inherited disease.

Thanks to 23andMe Founding R&D Architect Brian Naughton for his assistance in the preparation of this blog post.

Notes

1. Amerindian, sometimes just Amerind, is short for “American Indian”, and it denotes a descendant of the indigenous peoples of the Americas; anthropologist-types use the word to avoid confusion with the Indians that live in South Asia.
2. Because full-genome sequencing is still a few years from being affordable, researchers cannot look at every single one of the 3 billion nucleotides in the genome to find the one (or combinations of more than one) that are directly linked to a particular condition. For now, they must make do with the half-million to million SNPs that the current crop of SNP-genotyping chips allow.At first blush, this sounds like a fool’s errand. How can you possibly say anything about 3 billion DNA nucleotides with a collection of just a million markers? What if the marker you have on your SNP chip is near to but not actually the disease-causing SNP? Won’t this be like ships passing in the night? In just the last decade or so, geneticists have learned that the human genome has a very peculiar property: the 22 numbered chromosomes, or autosomes, and the X, tend to do the bulk of their recombining in a very small fraction of the spans of those chromosomes. These highly-recombining locations are termed hotspots; it’s kind of like chromosomes are trains, with hotspots as the links between boxcars (although unlike boxcars, which are all the same length, there’s considerable variation in the distance between hotspots). This state of affairs is good news for GWASes, because it means that these boxcars — big chunks (mean length ~200,000 DNA base pairs) of chromosome that tend to be passed from parent to child as a unit — aid the task of marker selection greatly. If you sample a bunch of people, as these researchers have done in Mexico, then you can build up a catalog of the specific chromosome chunks that occur. The idea is that when a disease-causing allele occurs, it can’t help but sit on one of these chunks, so the problem of building a GWAS chip is transformed into choosing markers that reliably distinguish between the chunks. It’s like being able to assign one inspector to each boxcar, rather than each crate inside it. All you need to do, then, is to look at SNP that’s diagnostic for the chromosome chunk that your disease-causing SNP is sitting on, instead of having to find the causal SNP itself. The technical word for these chunks or boxcars is haplotype³, and such catalogs when built are called haplotype maps; it’s actually what was meant by “medical genetic reference database” in the title of this post. It’s for the same purpose that the International HapMap Project, which has built reference haplotype maps for African, Asian, and European populations, was conceived, and it’s the origin of the name HapMap.
3. Haplotype is short for haploid genotype. Haploid means that you’re concerned with one chromosome, so haplotype means a contiguous segment from a single chromosome. Diploid means two (paired) chromosomes; humans are a diploid organism, because our chromosomes come in pairs. The fun doesn’t stop there. Some organisms are tolerant of higher ploidy, so there are tetraploid, hexaploid, and even octoploid species. For example wheat is hexaploid, so it has six copies of each chromosome.

ShareThis

Pop Quiz: What do women who eat cereal for breakfast each morning have in common?

1. They get a full day’s supply of 11 essential vitamins and minerals.
2. They enjoy better sex lives than women who don’t eat cereal for breakfast.
3. They are more likely to give birth to male children.
4. They make buying decisions based on the advice of cartoon leprechauns, silly rabbits and mustachioed mariners.

Actually, we don’t know anything about answers A, B and D. But we do know about C. When British scientists announced last year in the Proceedings of the Royal Society B that their research suggested eating breakfast cereal increases a woman’s chance of having a male child (among other things), they may have been falling prey to a statistical error that is of constant concern in the kinds of studies that 23andMe uses to provide information about the genetic components of disease risk. This post explains the nature of that misstep, and how we go about avoiding it.

Where did the researchers go wrong? Their research looked at the consumption of 133 different foods by 749 English mothers before and during the course of their pregnancies, and found that among all of these foods, only breakfast cereal was associated strongly with having sons. Nothing wrong with that. The problem arose, a pair of statisticians writes in the current online issue of the Proceedings of the Royal Society B, when the original researchers neglected to account properly for the number of opportunities they had had¹ to find a significant result.

In statistics, this is called the ‘multiple testing’ or ‘multiple comparisons’ problem, and it’s a very important concern for us at 23andMe. It’s pretty easy to get a feel for what can go wrong: suppose you were to flip a quarter five times, each time noting whether it came up heads. The odds that you would see heads all five times are not good — they’re 31 to one, or about 3%. But if you repeated this ‘experiment’ 133 times, you would be very likely to see all-heads at least once, and more probably three to five times. So if you see all-heads a few times, it doesn’t really make sense to regard this outcome as unusual. In fact, it would be very unlikely not to see all heads at least a few times². It’s something like the mathematical version of Tom Petty’s wistful refrain: Even the losers get lucky sometimes.

At 23andMe, the genetic association studies underlying our Health and Traits reports are saddled with the mother of all multiple testing problems. The basic idea behind these studies is pretty simple. They look at single-letter DNA variations known as SNPs in, say, 1,000 people who have a disease. Then they look at those same SNPs in another 1,000 people who are similar to the first thousand in as many ways as possible, except that they do not have that disease. If researchers were to look at a SNP and find, for example, that 80% of the people with the disease have the AA version, and only 30% without the disease do, then they’d be justified in suspecting that they’d found something significant.

But here’s the rub. These studies don’t look at just one SNP; they typically look at half a million to a million. A million is a lot more than 133; if association studies didn’t correct for multiple testing, every one of them would be expected to find tens of thousands of falsely associated SNPs mixed in with the really associated ones. That would not do.

Fortunately, there are many ways to correct for the multiple testing problem, ranging from dead-simple to math-degree-required. The top genetics journals will accept an association study only if the analysis has corrected for multiple testing using a standard method. Since we at 23andMe will consider a paper for use in Health and Traits only if it has come from a top journal (with some exceptions), we usually don’t have to worry about whether the studies we rely on are statistically sound. And, we go a step further by insisting that any finding used in Health and Traits must have been replicated by at least two independent groups of researchers. This policy helps to guard both against possible failure of multiple testing correction and unintended error on the part of the researchers conducting the study.

So how did the breakfast cereal-boy baby study make it into print? That remains a bit of a mystery to this reporter-slash-scientist. But anyone who is dreaming of having their own baby boy will just have to rely on alternative types of lucky charms.

1. The original paper looked at the womens’ diets before pregnancy and both in early and later pregancy, and tested each food in each time interval, so they appear to have conducted 399 (133*3) separate tests. They only report a significant association between consumption of breakfast cereal and bearing sons for the ‘before pregnancy’ and ‘later pregnancy’ time periods, so it is unclear whether they also tested the ‘early pregnancy’ time period. If they only tested ‘before’ and ‘later’, which seems odd, then that would mean they conducted 266 separate tests. Either number of tests, 266 or 399, suffices to cast serious doubts on the soundness of their result.
2. How unlikely? We want to know: what is the chance that in 133 5-coin-flip ‘experiments’, we never see all-heads? Well, per experiment, given a fair coin, the probability of getting five heads in five coin-flips is 1/32 (= (1/2)^5), about 3.1%. Therefore, the probability of not seeing all-heads is one minus the chance of seeing all-heads, or 31/32, about 96.9%. This is exactly the same as the chance of seeing at least one tail in five flips. The chance, then, that you see at least one tail in each of 133 flips is (31/32)^133, which is 1.47%. If you doubled the number of experiments to 266, the chance of never seeing all-heads (ie, seeing at least one tail in each experiment) is (31/32)^266, or 0.021%. This fits with your intuition: as the number of experiments goes up, the chance of never seeing an all-heads outcome goes down, because you have increased the number of opportunities in which to see this outcome.
   

ShareThis

This new feature, called Global Similarity: Advanced, shows you which populations from around the world you are genetically nearest to, giving you a clue as to where in the world your ancestors are likely to have lived over the last half-millennium or so.

Lilly Mendel in Global Similarity: Advanced’s World view.

How does it work? Let’s take 23andMe everywoman Lilly Mendel as an example.  In the image to the right she is shown as a green Google Maps-like cone on a background of colored squares, each of which represents a individual of known ancestry from our reference database.

The closer two people are on the plot, the more similar they are genetically. The genetic similarity is measured over all of Lilly’s autosomal SNPs, so it reflects the genetic contributions of all of her ancestors (unlike mitochondrial DNA, which is only inherited along the maternal line).

Lilly sits right in the middle of the European reference individuals (yellow squares), far from other reference populations like Native American and Central/South Asian; this means that Lilly’s ancestors were likely European. Lilly’s friends and family show up as black cones on the same graphic, so she can see where they each land with respect to the reference individuals and to each other.

World view shows our entire reference database at once. But Global Similarity: Advanced also lets you look at about a dozen further views, corresponding to different groupings of the world’s people. Lilly and her family are likely to be especially interested in the four European views, including Europe overall, and separate Northern, Southern, and Eastern European views.

Lilly Mendel in Northern European view.

Here’s an image of Lilly and the rest of the Mendel family in the Northern European view. The same reference individuals from World view are labeled here with their countries of origin so we can get a closer look at the Mendels’ ancestry.

The Mendels tend to fall closest to German and French reference individuals, suggesting that their ancestry is quite probably Northern European, and probably a mix of a number of European peoples.

Satisfyingly, one of 23andMe’s science team was born and raised in Dublin, and he falls right in the middle of the Irish cluster of reference individuals. Another 23andMe’er has one Eastern Asian parent and one European parent; she shows up about halfway between Europe and East Asia in World view. On the plot it looks like she’s closest to Central Asian populations, but it’s really saying that she’s roughly equidistant from Europe and Asia.

If the arrangement of the clusters of people looks a bit like that of a geographic map to you, 1) congratulations on staying awake in class, and 2) this isn’t an accident. As we’ve written recently in the Spittoon (here and here), geneticists are learning that genetic distance corresponds rather closely with geographic distance — or at least  it did until transoceanic ships and airplanes came on the scene. You can use your knowledge of geography to help interpret your results. For instance, our reference database currently doesn’t have any individuals of Korean ancestry. If you or a friend has Korean ancestry, it shouldn’t be a big surprise to find that person situated in the empty space between the Japanese and Chinese reference individuals, just as Korea is on a geographic map.

We’ve developed an animated Tour that both helps explain Global Similarity: Advanced and roughly documents the peopling of the Earth, beginning with the origin of modern humans in Africa about 200,000 years ago and their spread around the globe. Try clicking “Take a Tour” as you begin to explore the feature to start the animation. (And if you like that, be sure to give our fun and more detailed new video Human Prehistory: Prologue a try.)

The idea underlying Global Similarity: Advanced is simple enough. We calculate the genetic distances between all pairs of individuals that you’d like on the plot, again based on all the autosomal SNPs. In general, these points are arranged in some high-dimensional space, so you can’t readily picture them in two dimensions. We coax them down into two dimensions, like so: you lay all the individuals out in a plane, and move them around until the distances between each pair of points in the plane are as close as possible to the corresponding genetic distances. Once you find this optimal two-dimensional arrangement of the points, you’re done. There do arise some interesting technical issues when the rubber hits the road, and these are discussed in the feature’s white paper.

The reference database the feature is based on includes more than 1,200 individuals drawn from the CEPH-HGDP project, whose SNP genotyping 23andMe funded and made publicly available, and from Illumina’s iControlDB. This dataset does a great job at covering the breadth of human genetic diversity. Still, there are many peoples that we look forward to adding to the database to round it out, and in the near future we will be asking our customers if they wish to become part of this ancestry reference. As they say, watch this space.

Try out Global Similarity: Advanced for yourself or, if you don’t have an account yet, via one of our free demo accounts. Let us know what you think of the feature at help@23andme.com, or better yet, in the 23andMe Community!

ShareThis

As talks began Saturday at Cold Spring Harbor’s first “Personal Genomes” conference, the first half of which I blogged on here, several leading explorers of the strange new world of “structural variation” in the human genome, such as Evan Eichler and Mike Snyder, shared some of their latest findings.

You can observe the most common kind of structural variation by lining up two corresponding chromosomes; if you notice that one of them is missing a stretch of DNA letters that is present in the other, that’s a deletion (or an insertion from the other chromosome’s viewpoint). Geneticists call these types of variations insertion/deletions, or indels for short.

The main way these variations seem to happen is during the production of sperm and eggs, when an individual’s homologous chromosomes recombine with one another. Usually, chromosome pairs line up perfectly for recombination, but when they don’t, you end up with chromosomes that have a little more or a little less DNA than the originals.

Problems can arise when the stretch of DNA lost in the shuffle was part of an essential gene. Recent work has shown that one form of autism is caused by a deletion, and the same is true for forms of schizophrenia and cystic fibrosis.

Both basic research and clinically-motivated research were presented Saturday. Eichler, of the University of Washington, and Snyder, of Yale, are among those making detailed maps of structural variation in the human genome. The consensus is emerging that there is a *lot* of it. Snyder showed that, at least in the handful of individual genomes he’s looked at so far, any two individuals differ by more than one thousand indels of at least 3,000 DNA letters. Eichler made a similar observation with respect to his comparison between a sample individual’s sequence and the reference human genome, and commented that this means there are at least 3 million DNA letters (3,000 times 1,000) in the sample individual’s sequence not present in the reference genome. That’s on top of the 3 million one-letter SNP variants between any two humans. So it looks like genetic variation is the rule and not the exception, and the very notion of a “reference” sequence may need to stretch a bit in order to stay consistent with the data.

Derek Chiang of the Broad gave a fascinating talk about progress in trying to understand how structural variation can cause medical problems. Chiang, of the Broad Institute of MIT/Harvard, has developed sophisticated software to find structural variants associated with cancerous tissue based on microarray data. His efforts have already yielded a new oncogene for lung cancer. Chiang also described a clever algorithm for finding structural variants from next-generation sequence data that appears to work quite well by comparison to methods based on microarray data. Like the work Elaine Mardis presented Friday, Chiang’s talk suggests a future in which cancerous tumors could easily be distinguished from normal tissue using genetic scans.

The task of understanding what these differences *do* is another, very difficult, question entirely. And being able in turn to develop a therapy tailored to that specific change is yet another challenge.

Sunday’s talks took us into the world of *next next* generation sequencing, as though the prospect of plain old next generation sequencing weren’t already shaking up the field enough. Steve Turner gave a talk on Pacific Biosciences‘ sequencing technology that kept the audience rapt. The technology is ingenious, and solves a number of problems that have bedeviled sequencing since its inception. I will punt on explaining the basics of the technology, since it’s so well-explained on their website.

Turner offered an update that doesn’t appear on the website yet; the technology depends critically on the use of an enzyme, called DNA polymerase, that is responsible for copying DNA. Their team has modified the DNA polymerase used in the machine from the version that exists naturally in humans using a technique called experimental evolution. They generated a collection of mutant polymerases that each differ from the original at random. Then they tried each variant in their machine, retained the ones that do best, and repeated the process. It’s fair to say that horticulture and animal husbandry are slower, less direct forms of experimental evolution. After several generations, they ended up with a polymerase that was better suited to the environment of their machine than natural human polymerase is.

The meeting ended Sunday as Maynard Olson of the University of Washington, an eminent geneticist and one of the architects of the Human Genome Project, closed the conference with a witty and thoughtful summary of the proceedings. Olson suggested that true personalized medicine could be a long time coming, and expressed skepticism that it would arrive at all. One generalization from the talks, he said, was that it appears increasingly that the genetic mutations that impact health are rare, and it may be the case that many diseases can be caused in a large number of ways. The difficulty of working out what each version means could make it hard to work out therapies for so many different possibilities. He predicted a near-term future of unpredictability, especially in guessing which sequencing technologies will come to be adopted the genetics community. He closed by reciting the lyrics to Bob Dylan’s “The Times They Are A-Changin’,” which seemed fitting to me.

Photo: Henryk Kotowski

ShareThis

The structure of DNA was first publicly described 55 years ago at Cold Spring Harbor Laboratory (CSHL) on Long Island in New York by James Watson. Thursday night, the now 80-year-old Watson opened up the 2008 Personal Genomes meeting at CSHL by telling the story of the origins of the Human Genome Project, which he headed from 1990 to 1992. In 2003 the Human Genome Project produced a (nearly) complete reference DNA sequence of a human genome that is now essential to basic and applied human genetic research.

These days, Watson pointed out, scientists are able to read the DNA letters of the double helix so quickly and inexpensively that it is becoming practical to sequence the genomes of large numbers of people. With this progress comes a flood of research questions, technological challenges, and hope that these insights from the lab will translate into advances in personalized medicine.

Watson was followed on Thursday night by Francis Collins, who also followed him as director of the Human Genome Project, and later by Mary-Claire King, the renowned breast cancer geneticist from the University of Washington. Collins pointed out that health care costs have risen steadily over the years to the current level of roughly 20% of the US GDP. How much of this is spent on treatments that might have been identified as unnecessary with the availability of genetic information? He suggested that widespread genomic sequencing and analysis could lead to the discovery of the genetic causes of common diseases, such as lung cancer and Type II diabetes, for which some genetic links are now known, but much more remains to be learned.

Mary-Claire King considered breast cancer as a case study for personalized medicine. In the case of breast cancer, she noted, there are more than a thousand known mutations in each the genes BRCA1 and BRCA2 that can predispose a woman to the disease. Many of these are unique to specific families, or specific localities — she gave the example of one BRCA mutation endemic to a Norwegian valley. King illustrated through recent breast cancer studies that linking newly-discovered mutations to disease is a formidable technical challenge, but emphasized that the rewards for succeeding in doing so would be immense: roughly 5% of new breast cancer cases in the US each year — around 10,000 — are linked to known BRCA1/2 mutations, and thus might have been prevented through such measures as prophylactic mastectomy.

Friday moved into reports from the trenches. The morning session consisted of talks by researchers from major genome sequencing centers and from the companies behind the so-called “next generation” sequencing methods that underlie this conference. The new technologies, namely Illumina’s Solexa, 454’s FLX, and ABI’s SOLiD, follow the same general plan as the venerable Sanger sequencing method: scan short fragments, or ‘reads’, of DNA letters, and then reconstruct the original sequence from the reads. They just do it much faster than before, mainly by doing the scanning of many reads in parallel. Much of the concern these days is on the reliability of these new techniques — considering that a single changed DNA letter can mean the difference, for example, between getting Alzheimer’s or not — and so the presentations tended to focus on technical topics like error rates and comparisons across platforms. Even so, there were suggestions that some exciting new scientific findings might be around the corner; Richard Gibbs of Baylor showed early data from their sequencing of a HapMap trio (a father, mother and child) suggesting that the human mutation rate might be much higher than previously thought. And Elaine Mardis from Washington University showed that her lab had been able to find mutations unique to tumor tissue in a lung cancer patient. Known as somatic mutations, they had arisen in the patient during their lifetime, and were not found in non-tumorous skin tissue from the same patient. Her study did not show that one of these mutations had actually caused the cancer, but the demonstration that such changes may even be found is intriguing.

The afternoon session moved into the imposing task of storing, processing and interpreting the flood of data these new technologies generate. Paul Flicek from the European Bioinformatics Institute produced that rarest of things, the funny bioinformatics talk, in describing the travails of dealing with the 100 terabytes (that’s 100,000 gigabytes, or 100 million megabytes) generated so far by the pilot phase of the 1000 Genomes Project , and the specter of dealing with a petabyte (1,000 terabytes) of sequence data. Carlos Bustamante of Cornell described some of the insights into human evolutionary history that have made possible by the DNA deluge, including using sequence data to infer possibly the most detailed models yet of historical human population size and migrations. He also described his lab’s and John Novembre’s recent findings on the relationships between geography and human genetics; a topic we’ve blogged on recently at the Spittoon here and here.

There’s another big day of talks to come here at CSHL. I’m glad to be here keeping up to date on the latest research, so we can incorporate it into 23andMe, and to show off the site to a bunch of people on the cutting edge of genetics.

ShareThis

Now, 2,000 generations later, 23andMe is launching a new feature that can help you see the traces of those ancient voyages in your chromosomes. As populations became separated over the millennia, small genetic differences developed that can still be used as signatures of geographic ancestry. Ancestry Painting looks at those signatures in the 22 numbered chromosomes (that is, all but X and Y) to infer where in the world the ancestors who passed you each stretch of DNA were most likely to have lived – Africa, Asia or Europe.

Ancestry Paintings essentially give you a snapshot of where your ancestors lived before the beginning of the colonial era about 500 years ago. That’s because the massive migrations that have occurred since then have blurred many of the genetic boundaries that had developed over the millennia. For example, most Americans trace their ancestry not to the continent where they live but to Africa, Asia, Europe or a combination of those places.

The best way to understand Ancestry Paintings is to take a look at some. For example, this is the painting of a woman who has three grandparents from Germany and one from Norway:

Her painting is nearly uniformly navy blue, an indication that all of her chromosomes come from European ancestors. There are a few brief stretches of orange (meaning Asian ancestry), but in light of the overall pattern it is more likely that the orange stretches are statistical “noise” than indicators of true Asian descent.

Now suppose that woman had a child with an African man. Because each person has 22 pairs of numbered chromosomes – one from each parent – Ancestry Paintings depict both in a single band. So the child of a mother with fully European ancestry and a father with fully African ancestry would have a painting consisting of 22 half-navy and half-green stripes.

It’s in subsequent generations that things really get interesting. For example, this is the painting for a man who identifies himself as African American:

Genetic studies show that most African Americans have at least some European ancestry, and that finding is borne out by this man’s Ancestry Painting. In fact, all the chromosomes are at least half blue, suggesting that this person has one parent of fully European ancestry. The solid blue color in some stretches indicates at least partial European ancestry for the other parent, whereas the stretches that are half blue and half green indicate African descent from one parent and European from the other. The overall fraction of each type of ancestry – African, Asian and European – is tabulated in the legend in the upper right of the display.

Log into your 23andMe account to find many more examples of paintings from individuals from around the world, along with descriptions of the events in human history underlying each and more detailed information about how to interpret them.

ShareThis

If you take two members of the human race at random and ask how much their genomes differ, you’ll get a surprising answer: they’re almost identical.

On average, for every 1,000 DNA bases you have, 999 or so of them are exactly the same between you and your neighbor – and for that matter, between you and your neighbors on the other side of the planet. Over your entire 6 billion DNA base pair genome, however, that one difference in a thousand adds up to several millions of differences.

In studies published by Science and Nature this week, scientists have taken their most detailed look yet at these genetic differences. In this blog post, we’ll take a brisk stroll through their findings.

Both studies are based on around 600,000 SNPs genotyped in individuals (i.e., people) from the Human Genome Diversity Panel(HGDP-CEPH). HGDP-CEPH is a remarkable scientific resource. It consists of immortalized cell lines from 1064 individuals in 51 populations scattered around the globe. The idea guiding the creation of the panel was to take a wide-angle snapshot of human genetic diversity. This explains the presence in the panel of little-known populations like the Uygur, the Surui, and the Xibo, alongside more familiar populations like the Japanese, Palestinians, and French. This polyglot collection reposes at the Fondation Jean Dausset in Paris, as it has now for nearly a decade.

The Science study peers closely into those one-per-thousand differences between people, asking: Of all the genetic diversity seen in the panel, how much is found between people from the same population, how much between people from different populations in the same geographic region, and how much between people from different geographic regions?

For example, if there were no within-population diversity that would mean that all Russians are genetically identical, all Surui are identical, and so on, and therefore that all human genetic diversity must exist at the population and regional levels.

The paper finds nearly the opposite. About 90% of human diversity exists within populations, with most of the remaining 10% existing between geographic regions. This strongly confirms a decades-old result in human genetics: of those very few DNA bases which differ between people, a small minority of these differ between peoples.

Even so, 10% of several million differences is still a lot of differences between populations. Both studies zoom in on these differences, mustering some mathematical machinery called Bayesian cluster analysis, and ask: how easy is it to guess someone’s ethnicity based on their genotype? The answer the two papers find is that it’s pretty easy, at least on the regional scale. The following figure, drawn from earlier work done by many of the same researchers that was published in the open-access journal PLoS Genetics in 2005, illustrates the results of the analysis; these are qualitatively the same as the results shown in the fancifully-priced Science and Nature figures.

Each one of the (very) thin vertical lines in the figure represents a person, and the colors comprising each line correspond to the inferred proportion of ancestry from each of seven world regions. The key here is that the cluster analysis has no notion of a region or a population. It is simply told to divide up the genetic diversity into seven clusters (or six or eight – the results don’t change much), and then to guess which cluster or clusters each individual belongs to. The ethnic and regional labels are only applied once the analysis is through and, as is plain, the agreement between the donor-supplied ethnic label and the assignment is quite strong.

The papers go much further than we’ve seen here, looking into the history of human migration and exploring what happens when you use DNA insertions and deletions instead of SNPs to ask the same questions.

It’s worth noting that 23andMe is proud to have cosponsored the genotyping of the HGDP-CEPH that was conducted by the authors of the Science paper. The genotypes are available, gratis, at the CEPH website. We downloaded them ourselves, and now our customers can compare themselves to these very same populations using our Global Similarity feature.

ShareThis