Beyond the unfolding of complete genetic sequences lies the challenge of identifying and deciphering all the proteins that make up living organisms. Structural genomics—a new field catapulted into feasibility by the success of gene-sequencing projects and advances in the tools of structural biology—approaches that task through the large-scale determination of three-dimensional protein structures.

A protein’s genetic sequence can provide clues about its function, but a protein’s structure can better illuminate its biological action and its role in health and disease. A solved, high-resolution structure maps all the protein’s atoms, exposes surface topology and inner architecture, reveals electrochemical properties, and presents a testing ground for possible molecular partners. It paves the way for advances in structure-based drug design and the development of new medical devices and materials.

Determining high-resolution protein structures is often difficult and time-consuming, however. The essential tools of structural biology—X-ray crystallography and nuclear magnetic resonance (NMR) spectroscopy—each have their drawbacks. The former requires crystallization of the proteins, a laborious task, and the latter, though it uses proteins in solution, is usually slower and is limited to solving the structures of small and medium-sized molecules.

Jacob Keller, Columbia University

A protein of Methanobacterium thermoauto-trophicum, MTO146 is one of the first structures determined by the Northeast Structural Genomics Consortium and reveals information that could alter the protein’s original functional assignment. It is also known as precorrin-8W decarboxylase, a label that Columbia University (New York) crystallographer John Hunt, in whose lab the structure was solved, expects will change. "Enzymologists categorized the protein as a decarboxylase indirectly based on its sequence. We’re not sure if it’s a decarboxylase," he says, "but it’s unambiguously a methyltransferase" based on the group’s solved structure of the protein bound to S-adenosylmethionine. Recounts Jacob Keller, a research assistant in Hunt's lab, "I was simply playing around with the structure . . . superimposing structural homologs, when one of the homologs brought its AdoMet [S-adenosylmethionine] along, fitting it right into the homologous pocket in MT0146. Since all of the key contacts were in the right places, I co-crystallized the enzyme with AdoMet, and the new structure showed AdoMet density half an angstrom from where we expected it. It was very satisfying." Although they still don’t know all the details of the enzyme’s action, the group’s structural data have convinced them that the protein adds methyl groups to a vitamin B12 precursor. Previous work from other groups shows that the enzyme is present in all organisms that make vitamin B12. As an extra bonus, it contains a structural motif never seen before: a b-barrel tetramerization domain.

Structural genomics focuses on cranking out, at industrial speed, thousands of carefully selected structures from which most others can be predicted computationally with a reasonable degree of accuracy.

This approach relies on a belief in nature’s economy—that the countless different proteins in nature fold into a limited number of shapes and that all natural protein structures are a subset or combination of these shapes.

Andrzej Joachimiak, Argonne National Laboratory

Originally thought to look like a DNA-binding protein, this protein structure solved by Andrzej Joachimiak of Argonne (Ill.) National Laboratory, who leads the Midwest Center for Structural Genomics, turned out to be an enzyme with cyanase activity. The work illustrates how structural genomics can shed light on the evolution of protein function. This cyanase converts toxic isocyanide to ammonia and carbon dioxide, making it potentially useful as a detoxifying agent. It is a decamer composed of five dimers. The enzyme’s amino acid sequence is not similar to any other known protein, and apparently neither is its structure. "The subunits of cyanase are arranged in a novel manner both at the dimer and decamer level," according to Joachimiak.

The key to structural genomics is to group proteins into families of similar structures based on their sequences. Then, based on the known structure of at least one protein in a family and using a computational technique called homology modeling, a good guess can be made about the shapes of other proteins in the family. Estimates of the number of protein structure families range from 30,000 to 50,000—orders of magnitude smaller than the total number of proteins in nature.

Thinking Globally, Acting Locally

Currently, there is funding for structural genomics projects in the U.S., the European Union, Japan, China, Canada, and Israel. (In early April, representatives from four continents gathered in Virginia to discuss goals, progress, and policy issues; see "International Airing.")

Pharmaceutical companies and biotech start-ups are also committing to structural genomics, primarily to aid drug discovery.

The publicly funded U.S. effort is spearheaded by NIGMS, which last September launched the Protein Structure Initiative and will spend $150 million over the next five years on seven structural genomics pilot research centers, including one co-funded by NIAID (see "The First Seven."). NIGMS expects to fund a few additional centers this September.

These pilot centers will develop new techniques to streamline and accelerate every step in structural genomics, from choosing which protein structures to solve to cloning and purifying the proteins, determining the structures, and depositing the data into the Protein Data Bank (PDB), an online database of macromolecular structures, maintained by the Research Collaboratory for Structural Bioinformatics.*

In five years, each of the centers will ramp up to a production level of 100 to 200 structures annually at a significantly reduced cost per structure. Using traditional techniques, it takes weeks to months—and an average of more than $100,000—to solve the structure of a single globular, soluble protein. More recalcitrant proteins, such as membrane proteins, are even more challenging.

One long-term goal of the NIGMS project is to develop a public library of nature’s protein shapes that integrates sequence, structural, and functional information. This library should enable researchers to use genetic sequences to predict the approximate structures—and possibly the function—of any protein.

To build this public resource, NIGMS is enlisting its pilot centers to determine the structures of one or two representative proteins from each of thousands of different structural families. Ten thousand unique protein structures should be solved over 10 years, which includes the current five-year scale-up phase, then five more years at full speed.

Currently, of the 15,000 structures that have been deposited in the PDB, less than 4,000 are of unique proteins, defined as those whose sequences are less than 90 percent identical. And the solved PDB structures represent only about 1,500 families. By determining 10,000 protein structures from almost as many families, the Protein Structure Initiative would more than triple the number of unique structures available and would provide more thorough coverage of structural families.

One catch at this early stage is that there are many different ways to group proteins into families. The five-year pilot period should provide time to determine whether any particular method is better than the others.

The project also seeks to identify new folds. Proteins with the same fold have similar overall shapes but no detectable sequence similarity. Such proteins have the same types of structural components connected in the same order. Studying folds could reveal the physical and chemical principles that determine how proteins form their three-dimensional structures.

Scientists estimate there are only a few thousand folds–considerably fewer than the number of structure families—and only 700 of these are represented in the PDB.

Just Data Gathering?

In its early days, structural genomics was criticized by those who believed it was a rote exercise devoid of the creativity and intellectual challenge that characterize high-quality scientific research. Although such concerns are less common now, says John Norvell, who directs the NIGMS initiative, "it’s certainly true that structural genomics isn’t hypothesis-driven. It’s discovery-driven"—much like Darwin’s detailed observations and descriptions of finches, barnacles, and other creatures, which led to his theory of evolution, Norvell observes.

Although it is clearly too early to predict the eventual impact of the Protein Structure Initiative, like its predecessor, the Human Genome Project, it promises to open a whole new chapter in biomedical research.

* The Research Collaboratory for Structural Bioinformatics is a joint project of Rutgers University, Piscataway, N.J., the San Diego Supercomputer Center at the University of California at San Diego, and the National Institute of Standards and Technology, Gaithersburg, Md. It is supported by funds from the National Science Foundation.

Return to Table of Contents