Treating gene expression patterns as a searchable constellation of stars. The idea was too far-fetched to tackle two decades ago, when Chris Q. Doe was a postdoctoral fellow. He wanted to build an atlas showing the three-dimensional location of transcription factors involved in generating the fruit fly's central nervous system (CNS). Other atlases, however, covered simpler one- and two-dimensional patterns.

As proteins responsible for initiating gene expression, the many transcription factors in a fruit fly CNS correspond to different types of neurons. Making sense of these patterns could help scientists discover how brain cells develop and go on to form complex circuits.

Doe, now an HHMI investigator at the University of Oregon, gives credit for the idea of creating an atlas to his former postdoctoral adviser, HHMI investigator Matthew Scott, who is now at Stanford University. Scott had developed an atlas of Hox gene expression for Drosophila. He and his colleagues discovered that Hox genes, which determine where limbs and segments will form in the developing embryo, are expressed in a one-dimensional pattern. When Doe struggled to make sense of his three-dimensional data, he recalls Scott saying: "Just build an atlas."

Easier said than done. "I remember thinking it would be impossible—and that was with only a handful of CNS genes identified," says Doe. He dropped the idea as technologically unfeasible. The seed, however, was planted.

Now, 20 years and more than 200 transcription-factor genes later, Doe and his postdoc Michael Layden are on the verge of seeing such an atlas become a reality, thanks to a collaboration with computer scientist Eugene W. Myers, a group leader at HHMI's Janelia Farm Research Campus. Doe and Myers met at a neuroscience conference held at the Ashburn, Virginia, campus in March. "We've been working hard at this ever since," says Doe. "The methods of computational biology are now so advanced that it's conceivable to make a computer atlas of the nervous system and map the expression of hundreds of genes."

And if that atlas were searchable, he notes, it would potentially allow scientists to solve two important mysteries in neurobiology: how a stem cell becomes a particular type of neuron, and how neurons come together to form circuits that allow for all the complex processes governed by the brain. "Our best hypothesis is that neurons that make up a circuit share a transcription-factor code," says Doe. "Many labs are looking to identify these circuits. We're hoping our atlas will help them determine which transcription factors those neurons have in common."

Illustration: Jon Han

	PAGE 1 OF 2	Continue