Genetics moves on

Genetic building blocks

Making medicines match

Proteins first, genes second

Viktor Barski of the Russian Engelhardt Institute of Molecular Biology studies DNA data from a microchip that speeds sequencing of the human genome. Courtesy Argonne National Laboratory.

Genes exist to make proteins, which, ideally, help you stay healthy. But mutated genes may make proteins that make you sick.

The glut of new genetic information from the Human Genome Project is an embarrassment of riches: We know the sequences of thousands of genes whose functions are totally obscure. As sequence data accumulates, traditional means of assigning functions to unknown genes seem painfully slow.

Timothy Mitchison, a professor of cell biology at Harvard Medical School, is promoting a bass-ackwards solution with great potential benefits. Reasoning that genes exist to make proteins, he wants to find the proteins first and trace back to the genes.

This so-called chemical approach to genetics, now getting under way at Harvard's Institute of Chemistry and Cell Biology, reverses the traditional method of finding the role of genes. Mitchison, who co-directs the Institute with chemist Stuart Schreiber, says, "This technique moves the drug to the front of the process" and should work "even before the protein is known."

The trick, Mitchison says, is to find chemicals that interfere with proteins (which is how most medicines work) and use resulting changes in the organism to understand the protein's role. The technique, if successful, would not only identify unknown proteins, but also candidate drugs that affect them. It may also get closer to the interesting question -- what the gene and its protein actually do.

Traditionally, geneticists find disease genes by looking at families with lots of genetic illness. Eventually they "read" the gene and try to identify which protein it makes. Finally they look for a drug that affects that protein.

Before we get to Mitchison's brand-spanking-new wrinkle, remember that history proves that genes can be found through drugs. Willow bark, Mitchison observes, has been known to fight headaches since ancient times. Aspirin was extracted from it in about 1900, and in 1972, the enzyme inhibited by aspirin was identified, eventually leading to the discovery of the gene that makes the enzyme. Mitchison says that "classic piece of detective work" demonstrates how you can "Start with a drug, and work back to the protein and then the gene." Which is exactly what he wants to do.

But how to find molecules that affect proteins?

Good question (even if we asked it ourselves). The drug industry jealously guards its "libraries" of chemicals with drug potential, which academics can't afford to duplicate. But Mitchison says a new technique called split-pool synthesis allows a graduate student to make 1 million novel molecules in just a month.

Even for the average 24/7 grad student, that's serious output. Here's how it works: Say we've got a chemical core that can link to other chemicals at locations "A," "B" and "C." If three different chemicals can link at "A," four at "B," and five at "C," the simple math that even Why Filers understand says the possibilities total 3 * 4 * 5 = 60. (Most chemical cores can actually bind many more chemicals, but our math anxiety means you'll have to suffer our half-witted example.)

The opposite method also works. First you use chemicals that affect (bind to) a protein. Then you identify and analyze that protein. Finally, you determine the structure of the gene that makes that protein.

To start, a lowly graduate student attaches the chemical core to tiny plastic beads and exposes it to the various sub-units that can bind to location "A." Then the beads are mixed together, and the process is repeated with units that can bind to location "B" and then "C." At each step, the bead is coded to record what subunits it contains (this allows the chemical to be reproduced if it happens to do anything).

The new chemicals are then screened for activity in a soup of biochemicals. Those that are active (meaning they bind to a protein) may be tagged and used to isolate the protein. Then, by analyzing the protein's structure, scientists can predict what kind of gene sequence would produce it.

In a recent test run, the experimental technique created more than 2 million different molecules, about one-tenth of which have already been screened.

Makes drugs, too
Eventually, Mitchison hopes the technique could return the early steps of drug discovery from industrial to academic laboratories. It also upends a major route to drug discovery: For example, scientists found a protein called protease that allows HIV to multiply in AIDS, and then screened for chemicals that inhibit protease. These so-called protease inhibitors have revolutionized the treatment of AIDS and slowed the death toll in developed countries.

This giant freezer holds DNA. Courtesy National Institutes of Health.

In chemical genetics, we identify proteins by finding something -- call it a candidate drug-- that affects it. Thus the process should be attractive to pharmaceutical companies, Mitchison says, because "by definition you already have a drug -- and that could cut two years off the drug discovery process."

Certainly, there's plenty of room for improvement in the drug discovery field. While most drugs work by affecting proteins, all of the estimated 600 active ingredients in human drugs affect roughly 300 proteins. Mitchison figures that leaves roughly 200,000 human proteins waiting for drugs that might affect them.

And among that list we can expect to find proteins that influence just about every disease you can name.

Confused enough already, or do you want to read up on this stuff?

©2002, University of Wisconsin, Board of Regents.