Strength in Numbers

A mathematical mind has helped Leonid Kruglyak scan millions of yeast for the secrets of genetic complexity.

© Denise Applewhite for Princeton University

Leonid Kruglyak did his graduate work in physics, but when he dove into biology, he jumped with both feet. “The first thing I wrote about genetics was an eight-line letter to Nature,” he says. In it, he defended Dean Hamer’s 1993 discovery of a genetic basis for homosexuality—and took on Nature’s editor. “John Maddox had written this editorial basically trashing the Hamer paper, but he got the basic genetics wrong,” says Kruglyak. “You could have a lot of issues with the paper, but he suggested a model to explain the data that was completely inconsistent with the genetics. I was teaching myself human genetics at the time and I thought: ‘Wait a minute, the one thing it can’t be is this model...

“I don’t know what I was thinking,” he laughs. “Nature published two letters about that editorial. One from Dean Hamer and his collaborators. And one from this guy who was totally unknown.”

But he didn’t fly under the genomic radar for long. As a research scientist in Eric Lander’s lab at the Whitehead Institute, Kruglyak assembled programs that allowed investigators to do linkage analysis in humans. His algorithms, which formed the backbone of the popular program GeneHunter, “became the standard for mapping genes by their phenotypes,” says David Botstein of Princeton’s Lewis-Sigler Institute for Integrative Genomics.

And as an independent investigator, Kruglyak has turned his attention to model organisms, setting up experiments designed to tease apart the intricate interactions that underlie complex genetic traits in yeast and worms. “He’s really trying to understand how genes cooperate to produce complicated phenotypes,” says Botstein. “He’s not the first person to think this is an important problem. But in my opinion he’s been the first one to make any headway.”


For Kruglyak, genetics makes sense because it’s mathematical. “As soon as I started reading about it and thinking about it, it just connected with how my brain is wired,” he says. As a kid, Kruglyak read textbooks on astrophysics and calculus. “I worked out the binomial expansion on my own when I was 9 years old,” he says. “Then I started looking for it in books and was disappointed to discover that Newton had gotten there first.”

Kruglyak majored in physics as an undergraduate at Princeton and went on to develop physics-based models of neuronal networks and neuronal processing as a grad student at the University of California, Berkeley. Then he paused for some reassessment. “I did two short fellowships: one year at the Institute for Advanced Study and one year at Oxford,” he says. “The year at the Institute was figuring out that maybe I didn’t really want to do neurobiology. And the year at Oxford was looking for an alternative.”

The alternative he settled on was a puzzle with about 3 billion pieces. “It was 1992 and the genome project was beginning to gain visibility and papers were being published. So I started reading them,” says Kruglyak. And he started writing to the major players in the field. “Somehow I managed to talk my way into an interview with Eric Lander,” he says. Lander—who’d trained as a mathematician at Princeton and had also logged some time in Oxford—was a good match for Kruglyak. “Here was somebody who understood the math and who was really at the forefront of what was going on in genome biology and human genetics,” he says. “So it was an ideal environment.”

During his first year in the Lander lab, Kruglyak cranked out five papers, including a 1995 Nature Genetics piece on how to stringently interpret linkage studies. With nearly 3,500 citations, Kruglyak says, that rookie paper “is still the most cited thing I’ve ever done.”

Of course, his transition into biology sometimes required a bit of translation. “I remember one group meeting where Leonid got up and said—to a room full of biologists—‘I am now going to explain to you my method, using the fast Fourier transform as an analogy,’” recalls Bruce Hamilton of the University of California, San Diego. “I never saw Eric Lander laugh so hard.”

Over the next couple of years, Kruglyak helped build maps of the mouse and human genomes. He devised algorithms that allowed investigators to home in on their favorite genes. He even rolled up his sleeves and did some actual bench work, helping to screen a library for the gene that puts the stagger in staggerer mice.

“It was clear I wasn’t going to do that for a living,” Kruglyak says of handling pipettes and test tubes. But the experience gave him a better feel for how to analyze experimental data. And it stood him in good stead for launching his own biology lab when he joined the faculty at the Fred Hutchinson Cancer Research Center. “Not that I could seriously troubleshoot the technical details of difficult experiments,” he says. “But I do have some first-hand concept of what goes on in the lab.”

“It’s pretty unusual to have somebody that has a good eye for how to do a biological experiment and also really knows how to analyze data— really knows how to analyze data—instead of using other people’s methods and not really understanding them,” says David Goldstein of Duke University. “Leonid gets both things right. He designs good experiments and he sure knows his way around data analysis.”


When he first landed at the Hutch in 1998, Kruglyak continued to work on computation. His 1999 Nature Genetics paper predicting that the community would need to collect half a million SNP markers before it would be able to carry out comprehensive genome-wide association studies (GWAS) was widely read—but his conclusion was “extremely unpopular,” says Kruglyak. “Because a subset of the community just didn’t want to believe it.”

“Everybody was all excited about GWAS,” says Botstein. “And then Leonid comes along and says, ‘Wait a minute. This is going to be much harder than you think.’ And he was right.”

He was also starting to think more about how to get to the fundamental roots of genetic complexity: how to identify the players that interact to generate multigenic phenotypes. “That’s when I had the idea that it would be really productive to try to attack this problem in the simplest system in which these types of genetic interactions go on, but we still don’t understand them,” says Kruglyak. So instead of looking at human data, he decided to start raising yeast.

“Leonid’s approach is superb and unusual and right at the cutting edge.”
—David Altshuler of the Broad Institute and Harvard Medical School

“You might think that’s an obvious thing to do: use the most powerful model organisms to ask questions about complex traits. But remarkably, that’s generally not what’s been done,” says David Altshuler of the Broad Institute and Harvard Medical School. “So Leonid’s approach is superb and unusual and right at the cutting edge.”

The complex trait Kruglyak chose to focus on was transcription itself. “Gene expression levels vary from individual to individual,” he says. And the expression of any given gene is governed by multiple factors, including regulatory proteins and the sequences to which they bind. With funding from the Howard Hughes Medical Institute, Kruglyak set up a wet lab on the one extra bench he had in what was otherwise a space for computation. And he hired a pair of postdocs who crossed two yeast strains that had pretty different gene expression profiles. They collected 40 progeny and used microarrays to measure the activity of all their genes.

Then Kruglyak and company ran linkage analyses to determine how many genetic loci influence the expression of each gene. Some transcripts were controlled in a Mendelian fashion—that is, a single polymorphism in the gene or a regulator was responsible for the differences in its expression levels between the parent strains and their offspring. But more often, it appeared that a healthy handful of genetic factors influenced the expression of any given gene.

The resulting paper—published in Science in 2002—“spawned an entire field of research: using gene expression as a quantitative trait and being able to look at the inheritance patterns of thousands of traits simultaneously,” says David Gresham of New York University. “Being able to detect the complexity underlying quantitative traits that are segregating in natural populations in a single experiment is just a giant step forward.”

“The idea was pretty simple,” says Gaël Yvert of the Ecole Normale Supérieure de Lyon, one of those Kruglyak postdocs. “But doing the analysis was not so simple. We had to write every single piece of code to analyze the data because there was no software that could do what we wanted to do. And Leonid is good at that.”

“He’s also good at figuring out what the interesting problem is,” says Elaine Ostrander of the National Human Genome Research Institute, who was Kruglyak’s colleague and collaborator at the Hutch. Together, their labs conducted genetic analyses to determine how different dog breeds are related. “He’s got to be the only guy who has been a senior author on Nature papers about dogs and worms and yeast—all in a 5-year period,” she adds. “So that sets him apart. As does his sense of wonder about the natural world, which I think will keep his science fresh and unique and interesting for the duration.”

And the sheer size of the questions at hand will surely keep him busy. “Leonid is fearless about going in and doing the experiments at a scale that’s at least an order of magnitude beyond what anybody has ever done before,” says collaborator Patrick Phillips of the University of Oregon. Now at Princeton, Kruglyak has even expanded the scope of his original yeast experiment and is now doing linkage analyses on 10 million progeny at once—work published in Nature in April. The souped-up experiment allowed him to get closer to identifying all of the genetic factors that feed into a complex phenotype, in this case drug resistance. In some cases, the team was able to point to 20 loci involved in shaping resistance. “That was fun, because I think we’re getting to the point where we have a chance of answering the questions we’ve all been struggling with for a long time,” says Kruglyak. “It turns out that most phenotypes have a complex genetic basis. And until we can understand that the underlying genetics, we can’t address most of the really basic questions that Mendel would have wanted to understand, like how is height inherited in the human population? I want to know the answers.”

“Leonid is just joyful about learning new things,” says Altshuler. “And he doesn’t define himself as a computational person that doesn’t do experiments, or a human geneticist who doesn’t do yeast. He does whatever he needs to do to answer the question.”

Interested in reading more?

Magaizne Cover

Become a Member of

Receive full access to digital editions of The Scientist, as well as TS Digest, feature stories, more than 35 years of archives, and much more!
Already a member?