ACADEMIA

Discovery of new drugs requires the screening of thousands of compounds to identify hundreds of candidates that are winnowed to dozens of effective agents.

A few go on to preclinical development, the stage at which most drugs fail, and fewer still make it further to become the subjects of expensive and time-consuming clinical trials.

Oak Ridge National Laboratory researcher Jeremy Smith, working with University of Tennessee assistant professor Jerome Baudry, is using supercomputers to speed the process. "If drug candidates are going to fail, you want them to fail fast, fail cheap," Smith said.

In one day, a petaflops supercomputer -- which executes 1 quadrillion calculations per second -- can analyze 10 million drug candidates binding to a protein.

Within a decade, exaflops machines are expected to handle the interactions of 10 million drugs with 1,000 different proteins, allowing predictions of effectiveness and side effects.