"For each molecule, you can generate thousands of molecular descriptors," Hoflack says. Computer models of ADME properties are part of the ABCD suite. But once again, it is the way that the data can be quickly scanned by sophisticated users that is the software's real strength.

As is the case throughout the industry, the in silico or actual synthesis of large numbers of compounds is child's play. "We could actually imagine a billion compounds to be made," Hoflack says, "and anticipate how well they would do with all of these important parameters, and in the end select the 300, 400, 500 we feel would be most suitable."

These winnowed molecules can easily be lumped into chemical families, Hoflack says. But further prioritization is not always rigorous. "In the past," he says, "we probably would have selected those families we were most familiar with. It was a subjective selection. Today, we try to generate lots of information on each of these families and pick the ones we think are most likely to end up with a drug. We try to use much more information to make those selections."

"Where everybody is struggling is at the level of data mining, converting data into information, and into knowledge," Hoflack explains. "Where we are making our major leap is that we are building a system that can not only gather that information, but also give us the tools to interpret it. It's all about ADME. It's all about tox. All of these things we try to understand early on."

M.U.

Back to How to Spell Discovery