Guidance from an innovative computational approach could speed up the process and cut down the cost of new drug development, researchers from the University of Chicago Medical Center and Columbia University suggest in a study to be published in the February 2008 issue of Genome Research, available early online.

The researchers analyzed specific properties of the human genes and proteins that serve as targets for nearly a thousand FDA-approved drugs. They identified a number of characteristics that were common among successful drug targets–and especially common among high-revenue drugs.

"To make a good drug, you need to find a good drug target," said Andrey Rzhetsky, PhD, professor of medicine at the University of Chicago and senior author of the study. "Here we provide guidelines for more efficient target screening. When a drug company must decide which target to pursue among pathologic pathways, this could provide useful estimates of each target's expected success rate."

Gleevec, Prozac, Viagra: these successful drugs all target specific proteins and provoke a desired response. They reward the people who need them, as well as to those who invent, manufacture and market them.

But he development of a new drug is complicated and expensive, "a fusion of art and science," the authors note. It involves finding an accessible drug target and a molecule that binds that target as selectively as possible. Then it must trigger a desirable physiological change.

But "every highly visible success," the authors note, "rests on an iceberg of invisible failures." Since the estimated cost of developing a new drug ranges from $800 million to $1.2 billion, "information that helps only a little bit," Rzhetsky said, "can still be quite valuable."

What characteristics, they asked, distinguished the targets that eventually became the focus of such successful drugs? How are these genes or proteins different from the tens of thousands of less desirable targets?

To find out, the researchers looked at the relationships between 919 successful drugs, their human gene or protein targets, and the functional properties of those targets.

They found that most successful drugs are very precise; 62% of them have only one specific target.
           
But that's just a start. "A target molecule with an appropriate, 'drugable' structure is a necessary but not sufficient condition for success," write the authors. But the selection of a prospective drug target is "a complicated balance of many considerations,"

Additional properties that distinguish most good targets include:

  • Connectivity--how many otherproteins does the target protein interact with directly? Effective drugstargets--according to Rzhetsky and colleague Lixia Yao, from ColumbiaUniversity--interact with about nine other genes or proteins, which isabove average but not extremely high. This is enough connections to have asignificant impact but not so many as to multiply the risks of seriousside effects. The very-high-revenue drugs, the authors note, tend totarget genes and proteins with slightly lower connectivity.
  • Betweenness--how often does thisprotein serve as the shortest path between two networks of proteins?Successful drugs, they found, tend to bridge two of more clusters ofinteracting molecules.
  • Consistency--how much does thegene for this protein vary from person to person? Limited individualvariation is better, so the drug works the same way in most people.
  • Tissuespecificity--whereis the gene for the drug target expressed? In just one tissue, say brain,or skin, or all over the body? Ideally, the researchers found that thebest targets were genes or proteins primarily expressed in one specifictissue, so the drug treats the disease at its core, without interferingwith other, healthy tissues or organs, and thus causes fewer side-effects.Six tissue types, they found, were significantly "undertargeted"by pharmaceutical research: male reproductive tissues, embryonicstructures, skin, cartilage, bone and lymph.
  • Overlap--Finally,successful drug targets significantly overlap with disease genes--those inwhich a mutation can cause a specific disease, such as cysticfibrosis--and with essential genes, that are required for normaldevelopment.

"We found that genes associated with successful FDA-approved drugs have several properties at the network, sequence, and tissue-expression levels that significantly distinguish them from other human genes," the authors conclude. "Although the drug target-selection guidelines that we suggest cannot replace expensive experiments, they can help at the earliest stage of a drug-development project."

This work was supported by the NIH and the Cure Autism Now Foundation.