Abstract
There is great appeal, scientifically, financially and temporally, for the use of predictions in the drug metabolism and toxicology evaluations used, and in many instances, required, in chemical compound design and development. Indeed, there have been great strides in hardware and software development, as well as many scientific advances that have made predictions not only feasible but in many instances, accurate, when compared with data acquired from lengthy and costly human or animal studies. However, despite the wide array of available software capable of making predictions in the fields of ADME/Tox, the user needs to err on the side of caution when attempting to fit their data from a new structural entity into a database where the learning set may be somewhat or even grossly different from the new entity. There must be a great deal of thought given to what programs are to be utilized as well as to the robustness of the data being entered - there is still no replacement for the expression “garbage in, garbage out”. Further, decisions must be made as to whether the predictions made by the software are superior to other ways of gaining human predictions from in vitro and in vivo ADME/Tox information - if acquiring in vitro data is easier (high throughput) and more reliable, then maybe a new predictive software program is not the best answer, but it if is, we need to choose wisely. Acceptance of predictive software is difficult but it will, eventually, be well worth the effort. Until that time, the end-user needs to go into the prediction business with his or her eyes wide open as to the potential pitfalls and ramifications of an over exuberance of predictions. “However beautiful the strategy, you should occasionally look at the results”. Winston Churchill (1874-1965). British Statesman.
6