Predicting the activity and toxicity of new psychoactive substances: a pharmaceutical industry perspective

Predicting the effect that new compounds might have when administered to human beings is a common desire shared by researchers in the pharmaceutical industry and those interested in psychoactive compounds (illicit or otherwise). The experience of the pharmaceutical industry is that making such predictions at a usefully accurate level is not only difficult but that even when billions of dollars are spent to ensure that only compounds likely to have a desired effect without unacceptable side-effects are dosed to humans in clinical trials, they fail in more than 90% of cases. A range of experimental and computational techniques is used and they are placed in their context in this paper. The particular roles played by computational techniques and their limitations are highlighted; these techniques are used primarily to reduce the number of experiments that must be performed but cannot replace those experiments.