Predicting Toxicity Without Touching the Toxin

Primary Investigator: Urs Jans

It’s dangerous and expensive to test toxins. Special lab equipment is needed to protect researchers and using animals is expensive and controversial.

Far better to do the whole thing using computers. Powerful computing can elucidate the structure of a compound and predict how it would react inside a living body. A team at City College of New York has been using lab-bench chemistry to test various pesticide reactions, feeding their information into computers, and using the real-life chemistry to validate the computer predictions.

It has been working, said Urs Jans, an environmental chemist at CCNY.

“There are always new compounds that are being created,” Jans said. “In the context of a threat for the Army, they are worried they might be used either in the battlefield or to conduct attacks on water supplies or something like that,” he added.

Working with computational chemist Edward Hohenstein, who recently left CCNY, and biochemist Simon Simms, Jans has been testing organophosphate pesticides. They’re widely used in agriculture and are often not terribly dangerous to people – think malathion and chlorpyrifos. But sarin and other nerve gases are highly lethal and have been used in terrorist attacks. They work the same way – by interfering with neurotransmitters, which are the message-carrying chemicals needed for brain cells to communicate.

“We measured reactivity in the lab and then compared it to who what my colleague predicted, using computational chemistry,” Jans said. The goal is to use only computers, bypassing the need for animal testing or for anyone to handle the chemicals at all.

“If you have to quickly decide how bad something is, how toxic something is, this method can be very helpful,” Jans said.

Scientists have been using computers to predict how molecules will react for decades, but they are not always accurate. “One way that this has been done is if you know the reactivity of molecules that are very similar in structure, it is reasonable to assume that a molecule that is similar in structure has a similar reactivity. People would extrapolate this way,” Jans said.

“This works well if the mechanism (of action) is the same.” But that is not always the case, and there are two layers to predict when asking whether a compound interacts with human enzymes: how well the compound physically docks to an enzyme, and whether it then reacts.

As computers improve, it gets easier to predict what a chemical would do to people. “You don’t have to use toxic chemicals. It’s faster to do. You could even automate it,” Jans said. “Computers get better, they get faster and therefore you can include more into the models.”