Drugs and Poisons
Chemicals can have beneficial effects for living creatures—in which case they are called drugs—or they can produce toxic effects—in which case they are dubbed poisons. Most members of the former category also have properties in the latter: hence unwanted side effects of medicinal agents. One of the ongoing challenges in medicinal chemistry is to maximize the good effects and minimize the bad ones, resulting in a favorable so-called “therapeutic ratio.”
But what if the side effects of a useful medicine could be predicted in advance just by knowing its chemical structure? This isn’t strictly possible with current techniques, but a significant advance in foreseeing adverse drug reactions is reported by a Harvard group using sophisticated statistical modeling.
The authors first assembled a group of 809 drugs with 852 known side effects and looked for associations between intrinsic drug properties and associated overlapping toxicity. Then they added 16 types of descriptive data (melting point, charge, bond types, etc.) and ended up with a highly accurate network for predicting unwanted reactions of new drugs that might not be detected in standard clinical trials.
This is practical stuff. While the only true verification of dangerous side effects is actual experiences in patients, avoiding the next Avandia, Vioxx, or Thalidomide would provide a tremendous boost to patient safety—not to mention a tremendous reduction in the liability facing those who would develop new agents intended to help humanity.
Tom Tritton is President and CEO of CHF.
Pharmaceuticals and Medicinal Chemistry [Discover CHF]