Legitimate estimates suggest that developing a novel chemical entity (NCE) as a drug can cost up to U.S.$ 2 billion [1,2]. Still, about 10% of NCEs show serious adverse drug reactions (ADR) after market launch [3]. The majority of these ADRs can be avoided if possible undesired off-target effects of the compound are understood very early during the drug discovery process, that is, before clinical trials are started.This contribution focuses on computational methods that are used to assist and to guide in vitro preclinical safety pharmacology (PSP), a technology commonly applied in the pharmaceutical industry to evaluate compound selectivity profiles [4][5][6][7][8]. To develop compounds highly selective for a therapeutically relevant target and to avoid side effects or adverse drug reactions are key goals for every small-molecule drug discovery project. To achieve this, preclinical safety pharmacology approaches are commonly employed to screen compounds routinely in comparatively inexpensive, yet predictive assays to generate knowledge about possible polypharmacology. Thereby a comprehensive identification of possible liabilities can be achieved.We outline the currently available environment and approaches that can be applied for thorough computational analyses of in vitro safety pharmacology data. After discussing desirable and necessary prerequisites for the data input from a computational perspective, we address how this data is used to predict a general promiscuity score for a single compound. This approach aims to answer the general question of whether a compound will hit many targets or will be selective. Finally, we demonstrate how to computationally reveal possible single-target liabilities. This has the objective of understanding why a certain compound is active against a defined undesired target on a molecular level.The above approaches are collectively used to triage compounds prior to being screened in an in vitro safety pharmacology panel in order to prioritize compounds for testing and identify those which have the highest likelihood of being selective.Hit and Lead Profiling. Edited by Bernard Faller and Laszlo Urban