Chemistry Reference
In-Depth Information
In silico
techniques for the prediction of toxicological endpoints are highly
desired in the drug discovery industry because of their fast return of results,
inexpensiveness, the advantage of identifying problematic candidates and
evaluation metabolites safeness before synthesis [43-46].
Toxicity is responsible for many compounds failing to reach the market and for
the withdrawal of a significant number of compounds from the market once they
have been approved. In 2003, it was estimated that approximately 20 to 40% of
drug failures in investigational drug development could be attributed to toxicity
problems. Five years later, toxicity was still the cause of 20% of the dropouts
during late development stages [3, 44].
ADME/Tox
in silico
models can fail due to the expectations of the users or due to
the development aspects of the model, such as choice of statistical tool,
description of the investigated structures and limited model validation [40]. To
develop predictive
in vivo
toxicity models based on
in vitro
or
in silico
inputs,
large, high-quality collections of animal data are required for training and
qualification, and unfortunately much of the world's
in vivo
animal testing data is
widely dispersed [42].
There are basically two main
in silico
approaches commercially available to
forecast potential toxicity. One approach uses expert systems (ES) that derive
models on the basis of abstracting and codifying knowledge from human experts
and the scientific literature. The other approach relies primarily on the generation
of descriptors of chemical structure and statistical analysis of the relationships
between these descriptors and the toxicological end-point, such as carcinogenicity
[3, 44]. Benfenati and Gini [38] classified the toxicity preview programs into three
categories: Rule-based human-derived ES (DEREK), ES using statistical
procedures (TOPKAT and CASE) and ES encoding mechanistic processes
(COMPACT).
DEREK (Deductive Estimation of Risk from Existing Knowledge) is the example
of an approach based on human expert's knowledge, which evaluates an unknown
chemical looking at its similarity with other molecules. It is a qualitative tool used
in agrochemical, pharmaceutical and regulatory organizations. DEREK also takes
