Machine learning and access to ever-expanding databases improves genomic prediction of human traits.

In theory, a scientist could predict your height using just your genome sequence. In practice, though, this is still the stuff of science fiction. It’s not only your genes that affect height—environment also plays a role—but the larger problem is that height is affected by tens of thousands of individual genetic variations. This is also true of other complex traits, such as susceptibility to particular diseases. To get closer to accurate genomic prediction of human traits, geneticists are using new approaches to harness the vast amounts of sequence data becoming available. In GENETICS, Lello et al. describe a machine learning approach to the problem that allowed them to make predictions within a few centimeters of reality.

“To me, genomic prediction is the actual decoding of the genome,” says senior author Stephen Hsu from Michigan State University. A theoretical physicist by training, Hsu explains that his lab became interested in the problem of genomic prediction several years ago as the cost of genotyping continued to drop and more datasets became available. They had previously argued that they could predict complex traits, like height, if they only had enough data.The release of nearly 500,000 UK Biobank genotypes allowed them an opportunity to test this hypothesis.

A genomic prediction approach is quite different from the more familiar genome-wide association study (GWAS). GWAS methods test each SNP one at a time, looking for statistically significant contributions to the phenotype. In contrast, genomic prediction makes use of all SNPs at once in trying to build the best possible predictors.

The authors took the Biobank genotype and phenotype data and used a type of regression to identify the combination of SNPs that, taken together, best correlate with the trait of interest. Since only a subset of SNPs influence each trait—even the thousands of loci that control height are only a tiny fraction of the total number of SNPs identified —they also introduced a penalization factor that prevents the model from including unneeded SNPs. They were essentially trying to solve an optimization problem: identify the fewest number of variables (i.e. SNPs) that will allow for the best prediction about the outcome (i.e. trait).

Having generated their algorithm, the authors then put it to the test. They constructed models for height, heel bone density, and educational attainment, and they found that their algorithm worked well, particularly for height. For example, it produced a nearly 0.65 correlation with actual height, and predicted heights were usually within a few centimeters of actual heights. “Our predictor actually captures almost all the heritability that we could expect to find,” says Hsu.

With enough data, Hsu believes, accurate genomic prediction for complex traits will no longer be sci-fi. As more and more genotypes are obtained, Hsu predicts that this kind of prediction could be applied for most traits in as little as five years.

CITATION:

Accurate Genomic Prediction of Human Height

Louis Lello, Steven G. Avery, Laurent Tellier, Ana I. Vazquez, Gustavo de los Campos, Stephen D. H. Hsu

Genetics October 2018 210: 477-497; https://doi.org/10.1534/genetics.118.301267

http://www.genetics.org/content/210/2/477

A new collection of inbred flies provides a tool for studying genetic control of sleep.

Sleep is vital for a healthy life, but some of us seem to get by with less snoozing than others. This individual variation isn’t unique to humans; fruit flies also show a variety of sleep patterns. These differences could potentially reveal more than just which flies are consistent cat-nappers—understanding the genetic basis of sleep variation could help pinpoint some of the molecular mechanisms that govern this essential and evolutionarily conserved process. In G3: Genes|Genomes|Genetics, Serrano Negron et al. describe a collection of inbred fly lines with extreme sleep behaviors that will serve as a useful tool for exploring such questions.

In a previous study, the authors had worked with flies from the Drosophila Genetic Reference Panel (DGRP), which is a large panel of inbred lines derived from a natural population. They chose the five longest-sleeping and five shortest-sleeping DGRP lines and allowed them to cross at random for 21 generations to produce an outbred population. They then used artificial selection to produce two long-sleeping and two short-sleeping populations.

In the new report, the authors created inbred lines from these long-sleeping and short-sleeping populations. Creating inbred lines is useful for gene editing studies because it reduces experimental noise caused by background genetic variation. To create lines of flies that sleep for either a very long or very short time, they mated one male and one female from one of the selected populations and then selected one male and one female from the progeny to propagate the line. They repeated this for 20 generations, eventually creating 39 inbred lines (19 long-sleeping lines and 20 short-sleeping lines), which were termed the Sleep Inbred Panel.

The duration of night sleep in the inbred lines ranged from just over an hour to nearly 11.5 hours, confirming that the extremes of the phenotypes had been maintained in the lines. The phenotypes were comparable to the selected parental populations, showing that the inbreeding process reduced genetic variability without drastically altering the sleep patterns. Interestingly, most of the variation between the new panel and the parental populations was due to short-sleeping inbred flies sleeping a little more than parental flies, which is likely because short sleep times made the flies less fit. Individual short-sleeping flies also showed more variation in how much they slept, likely for similar reasons.

By sequencing the genomes of flies in the panel, the authors identified a number of SNPs and other genomic variations associated with the sleep phenotypes, and they traced many of these variations back to the original DGRP lines. This resource can now be used to study the genetic underpinnings of sleep and may one day help shed light not only on ordinary differences between night owls and early birds but also on the causes of devastating sleep disorders.

Citation:

The Sleep Inbred Panel, a Collection of Inbred Drosophila melanogaster with Extreme Long and Short Sleep Duration

Yazmin L. Serrano Negron, Nancy F. Hansen, Susan T. Harbison

G3: Genes, Genomes, Genetics September 2018 8: 2865-2873;

https://doi.org/10.1534/g3.118.200503

http://www.g3journal.org/content/8/9/2865

Tissue culture causes heritable methylation changes in plants.

Tissue culture is a useful tool for plant scientists and horticulturalists in large part because it allows them to produce clones. Inconveniently, however, these clones are not always identical to the original, as one might expect them to be. In a report in GENETICS, Han et al. examined how propagation by tissue culture induces heritable epigenomic changes in maize.

When a portion of a plant is grown in tissue culture, it de-differentiates into an amorphous callus. This deprogrammed tissue can be induced to form roots or shoots or even to regenerate an entire plant—but this complex process can leave its marks on the genome and epigenome of the progeny. To get a picture of how tissue culture affects the epigenome, the authors compared methylation patterns in parental plants, plants that had been cultured, and the progeny of those cultured plants.

They found that most methylation was highly stable; it was consistent among all plants and unaffected by culturing. However, a subset of the methylome was variable between cultured and uncultured plants.  Many of these DNA methylation differences were passed on to the progeny of the cultured plants. Importantly, some of the changes the authors identified were shared among independently regenerated progeny, suggesting that tissue culture can prompt consistent, heritable epigenetic effects in maize.

In theory, these epigenetic changes might be due to general stress; for example, the culture process might cause the methylation machinery to become dysregulated. However, since most methylation in the genome was largely unaffected, and many changes were consistent among cultured plants, it’s more likely that these changes are targeted, with certain alleles being more sensitive than others to heritable epigenetic changes during culture. The mechanisms that lead to methylation modifications and the genetic and phenotypic consequences of those changes will be interesting avenues for further study; however, since most plant genome editing requires a culture step, researchers should be cautious about unintended epigenetic consequences.

CITATION:

Heritable Epigenomic Changes to the Maize Methylome Resulting from Tissue Culture

Zhaoxue Han, Peter A. Crisp, Scott Stelpflug, Shawn M. Kaeppler, Qing Li, Nathan M. Springer

GENETICS August 2018 209: 983-995; https://doi.org/10.1534/genetics.118.300987

http://www.genetics.org/content/209/4/983

Although epigenetic modifications contribute to trait variability, their effect pales in comparison to standing genetic variation.

The raw material of evolution is genetic variation, but proponents of the “extended evolutionary synthesis” add a new layer to this model: heritable variation in epigenetics. The packaging and tagging of DNA can alter traits without changing the DNA sequence, and in some cases, these changes can be inherited across generations. Can this epigenetic variation play a role in adaptation? Though this question is still under debate, a report published in G3: Genes|Genomes|Genetics suggests that the influence of epigenetic variation on trait variability may be comparatively feeble.

Aller et al. set out to directly compare the influence of genetic and epigenetic variation on an adaptive trait in the flowering plant Arabidopsis thaliana. To do this, they used epigenetic Recombinant Inbred Lines (epiRILs), which are bred from closely related plants with and without a specific mutation in a gene important for maintenance of DNA methylation, such that almost all of the heritable variation in their progeny is attributable to differences in which parts of the genome are methylated— i.e. epigenetic variation. The lines are essentially genetically identical, but each has a different stably-inherited pattern of DNA methylation.

For each epiRIL, the authors measured adaptive traits such as flowering time and accumulation of glucosinolates, which are compounds the plants produce for defence against herbivores and pathogens. The team then compared the variation in this epigenetic system to other studies that investigated the genetic variation underlying those same traits.

Although the authors found significant variation within their epigenetically-driven model, it was much lower than variation in genetically-driven equivalents. This suggests that epigenetic changes are much weaker drivers of variability than the major engine of adaptation: alterations of the genetic code.

CITATION:

Comparison of the Relative Potential for Epigenetic and Genetic Variation To Contribute to Trait Stability

Emma S.T. Aller, Lea M. Jagd, Daniel J. Kliebenstein, Meike Burow

G3: Genes|Genomes|Genetics 2018 8: 1733-1746. DOI: 10.1534/g3.118.200127

http://www.g3journal.org/content/8/5/1733 

Archaeologists have long known how to extract millennia-old stories from a single tooth buried in an ancient ruin—and now geneticists have the tools to join them. Advances made in the last several years have enabled researchers to sequence tiny amounts of DNA preserved in very old specimens, such as the material inside a tooth from the Stone Age. But this ancient DNA (aDNA) is often severely degraded, limiting its use. In GENETICS, Joshua Schraiber describes a new statistical approach to getting the most from these old samples and reports how he used the method to uncover secrets about the relationships between ancient humans and modern ones.

A major obstacle to understanding humans’ recent evolutionary history has been the inability to infer much about it using genetic data from people living today. If restricted to data from modern people, we would be locked out of information of great scientific and cultural relevance. The genetic relationships between ancient and modern populations can provide clues about migrations that occurred thousands or tens of thousands of years ago and help us better understand our histories. For example, researchers recently found that many people living in South America today are in part descended from an ancient North American group called the Clovis people.

Solving these kinds of puzzles is where aDNA shines—if you know how to use it. Schraiber found that the best way to determine genetic relationships among modern and ancient populations using degraded samples of aDNA is to sequence multiple ancient samples at low coverage rather than fewer samples at high coverage. After applying his new method to existing genetic data from 230 West Eurasian people who lived 8500–2300 years ago, Schraiber discovered that none of them came from populations that are direct ancestors of any modern European populations tested.

Schraiber’s analysis also suggests many ancient European people separated into small populations with little gene flow among them, and that most of these local groups died out, leaving a limited genetic legacy in modern European people. His results further imply that the oldest populations were the smallest, although this must be tested further because of the complicated nature of drift time, one of the parameters used in the analysis. Since drift time also complicates other types of analysis, Schraiber anticipates that methods to fully account for the troublesome variable will allow development of a fuller picture of these results.

If validated, the increase in effective population size over time predicted by Schraiber’s method would be interesting to compare to archaeological information about ancient humans, especially because many current ideas hinge on a link between agriculture and the rise of larger, more interconnected societies. Perhaps new scientific evidence would put to rest some debates about ancient humans’ lives—but it would surely spawn even more questions, too.

CITATION:

Joshua G. Schraiber. Assessing the Relationship of Ancient and Modern Populations.
GENETICS, 208(1), 383-398.
DOI: 10.1534/genetics.117.300448
http://www.genetics.org/content/208/1/383

For the past several decades, Shetland ponies’ collective past had caught up with them. A portion of the population of these miniature horses is affected by atavism, a phenomenon in which ancient characteristics are accidentally revived by mutations. Traits reincarnated in this way sometimes interact disastrously with the genetic background of the modern organism. For the ponies, a type of skeletal atavism (SA) has tragic consequences: the affected animals are born with some of the bones in their legs fused together, resulting in leg deformity and difficulty moving. Over ten percent of the ponies are carriers, and most affected individuals have to be euthanized at a young age. In an article in the July issue of G3, researchers report not only the cause of this condition, but also a genetic test that can be applied to breeding programs immediately.

Ponies suffering from skeletal atavism have bowed legs. Photo credit: Ove Wattle.

The researchers first attempted a genome-wide association study, which looks for statistical associations between a disease and variation at a region of the genome, but this approach failed to detect the mutation responsible for SA. To dig a little deeper, they then resequenced the whole genomes of six affected ponies and compared them to the genomes of a pool of control horses—stallions that had at least 50 registered offspring, none of which had SA. They found that the SA-affected ponies suffered from at least one of two partially overlapping deletion mutations, which they labeled Del-1 and Del-2.

These deletions were in sequences unassigned in the Thoroughbred horse reference genome assembly (EquCab2.0), which explains why the genome-wide association study originally failed to find them. These mutations fell in the pseudoautosomal region of the sex chromosomes, a stretch of DNA with a high GC and repeat content that makes it notoriously difficult to sequence and assemble—even in the human genome, about 600 kilobases remain unassembled. The region is very fragmented in the EquCab2.0 genome assembly, and the researchers’ attempts at de novo assembly using short sequence reads from the SA-affected ponies’ genomes also failed.

To circumvent this problem, they used single-molecule real-time (SMRT) sequencing, which enabled them to sequence longer stretches of DNA and pinpoint the deletions to the gene SHOX and the downstream gene CRLF2. Both Del-1 and Del-2 mutants were missing the region downstream of the gene SHOX, which includes the gene CRLF2, but Del-1 mutants were also missing the entire coding region SHOX. Taking into account their functions alone, either gene could plausibly be involved in the deformity. Among the genes’ many roles, CRLF2 is involved in bone metabolism and SHOX is developmentally regulated and activates transcription in cells that develop into bones.

An X-ray image of an atavistic pony’s legs. Photo credit: Göran Dalin.

Although both mutations completely deleted CRLF2, the researchers hypothesize that the effects on SHOX function are the mostly likely cause of SA—whether through deletion of the entire gene or loss of regulatory elements its downstream region. The apparent role of SHOX in humans supports this idea. In humans, haploinsufficiency or deficiency of SHOX is associated with limb deformities, and a 47 kilobase deletion downstream of SHOX is responsible for the skeletal disease Leri-Weill dyschondrosteosis as well as idiopathic short stature. While the results don’t suggest a cure for these human conditions, they did allow the researchers to develop a genetic test for SA in ponies, which—if applied—could allow this crippling condition to be bred out immediately.

CITATION:

Rafati, N.; Andersson, L.; Mikko, S.; Feng, C.; Raudsepp, T.; Pettersson, J.; Janecka, J.; Wattle, O.; Ameur, A.; Thyreen, G.; Eberth, J.; Huddleston, J.; Malig, M.; Bailey, E.; Eichler, E.; Dalin, G.; Chowdary, B.; Anderssson, L.; Lindgren, G.; Rubin, C. Large Deletions at the SHOX Locus in the Pseudoautosomal Region Are Associated with Skeletal Atavism in Shetland Ponies.
G3, 6(7), 2213-2223.
DOI: 10.1534/g3.116.029645
http://www.g3journal.org/content/6/7/2213.full

Scientists are known for being critical thinkers, experimental experts, and data enthusiasts. It’s probably no surprise that many of us are also undeniable nerds. Eric Spana, Assistant Professor of the Practice in Biology at Duke University and long-time GSA member, is no exception.

“We all have some type of nerd-ism, whether it’s Harry Potter, Marvel, Shakespeare, or sports,” he says. “Everyone is a fan of something a little more than they should be.”

Eric Spana

For the past two years, Spana has been combining his love of science and his nerdiness by giving presentations in the science track at Dragon Con – a multi-media pop culture convention that happens every Labor Day in Atlanta, GA, and attracts over 60,000 attendees.

During these presentations to packed rooms, he discusses the application of scientific facts to some of our favorite fictions. In 2015, he gave a talk on the genetics of the wizarding trait in the Harry Potter series. In the world of Harry Potter, witches and wizards can perform magic, which Spana defines as “breaking the rules of physics by just thinking about it.” Those who can’t perform magic (so, you and me) are Muggles. Two Muggles can produce a wizarding child, and two wizards can produce a non-wizarding child, known as a Squib, although this is rare. Generally, all members of a wizarding family can do magic. Author J.K. Rowling has stated that the wizarding gene is dominant, which – if true – requires a bit of explaining.

Spana says, “I look at it as engineering problem solving. If a writer creates something, it’s my job to figure out how to make it happen. If J.K. Rowling wants the wizarding gene to be dominant, then okay, I can make that happen.” Wizarding genetics has received quite a bit of attention among scientists and fans, with papers being published and fan theories being sent to Rowling herself.

Spana first read the Harry Potter books a number of years ago after his mother-in-law gave the first book to his daughter. “That was the beginning. We did the whole midnight Deathly Hallows thing. I’ve got Slytherin bag tags on my luggage and a Slytherin polo that I wear for exams, and it scares the students.”

He became interested in the genetic transmission of the wizarding trait while teaching a class on genetics and developmental biology. The beginning of the class covered basic topics like ploidy and recessive versus dominant traits, which Spana admits can be a bit dry.

“One semester, I had an unusually small class, and I could just see their eyes glazing over while we were talking about fruit flies. For some reason I made a Harry Potter joke, and they perked up. So we spent the next hour working through the wizarding gene, is it recessive or dominant, etc. The students piped up and produced answers because they know this stuff like the back of their hands.”

Since then, he’s refined his theory of a dominant wizarding gene through literature review, hallway conversations with colleagues, and discussions with an independent undergraduate researcher in his lab. His current explanation is that the wizarding trait is indeed dominant, with de novo mutations accounting for Muggle-born witches and wizards.

“You can calculate the mutation rate based on the number of live births in the UK and the number of Muggle-borns at Hogwarts. It comes out to be roughly 1 in 750,000, which makes it extremely rare. So, why didn’t you get an owl on your eleventh birthday [as all wizarding children do]? Well…statistics and probability!”

As for explaining Squibs, he’s settled on his colleague’s idea of a second-site transcription factor that inhibits the wizarding gene.

“The nice thing about the second-site transcription factor theory is that there is a little evidence that you could remove the Squib trait in the next generation. There was a character named Mafalda that was written out of Goblet of Fire, and she was the daughter of a Squib and a Muggle.”

Spana’s presentations at Dragon Con have gone very well, with standing-room-only audiences in large rooms of the convention.

The crowd for Spana’s 2015 Genetics of Wizarding talk.

“There were people in line for over an hour to see it, and they still turned people away! I’m just blown away that so many people would come see a college professor talk about genetics. That’s pretty awesome.”

He says that people love science, and as he explains recessive and dominant traits using slides illustrating the famously redheaded Weasleys and Harry’s redheaded mother and daughter, he can see the topics click for his audience.

“It really helps transmit genetics to a lay audience who already know the [fictional] material. What’s great is that this is outreach, but I also get to talk about dorky things I like. I think it’s a better way to do outreach. It’s taking things that people already know inside and out and giving them a framework to think about it a little differently.”

Savvy audience members have asked questions about other interesting members of the wizarding world, including Nymphadora Tonks, a Metamorphmagus (meaning she can change her physical appearance at will) who had a Metamorphmagus son but was born of non-Metamorphmagus parents. There are giants and near-human magical peoples (like Veela) to account for, and the very interesting subject of population genetics is ripe for investigation.

Spana, ever the problem solver, is looking forward to tackling these subjects in future talks. And since Muggles can’t make potions (no magic, no wand, no potion!), Spana is just going to have to use science to explore the magic.