CHEMOINFORMATICS AND BEYOND
Catrin Hasselgren, Daniel Muthas, Ernst Ahlberg, Samuel Andersson, Lars Carlsson, Tobias Noeske, Jonna Stålring, Scott K. Boyer · 2013
Preclinical and clinical safety is one of the major reasons for attrition in the drug discovery and development process. To try and tackle this challenge in an efficient manner, experimental work is being accompanied by an increasing amount of computation and prediction. Since safety issues are often based on complex mechanisms and safety data come in multiple forms and levels of quality, multiple computational techniques are required to assist in this process. Each computational method carries with it specific benefits and drawbacks. In this review, we discuss how different methods, ranging from classical structural alerts and QSAR models to weight of evidence techniques, and more informatics-based approaches complement each other. We go on to outline how they are used at different stages in the development process driven by the available data and specific question. We also address an often-overlooked aspect of modeling: that successful implementation of these methods requires a stable and flexible computational system to bring the models to the users. It is clear that the amount of data that can be leveraged to make better safety-related decisions is growing rapidly. The challenge is to use the data in a way that exploits it fully without overinterpretation to make the output of modeling and informatics efforts comprehensible to the end-user and finally to assure that the entire infrastructure is stable and does not introduce its own set of problems and errors. Each of these aspects will be addressed specifically in this chapter.