Adverse drug effects are unintended and undesirable effects of medicines, causing attrition of molecules in drug development and harm to patients. To anticipate potential adverse effects early, drug candidates are commonly screened for pharmacological activity against a panel of protein targets. However, there is a lack of large-scale, quantitative information on the links between routinely screened proteins and the reporting of adverse events (AEs). This work describes a systematic analysis of associations between AEs observed in humans and bioactivities of drugs while taking into account drug plasma concentrations. In the first chapter, post-marketing drug-AE associations are derived from the United States Food and Drug Administration Adverse Event Reporting System using disproportionality methods, while applying Propensity Score Matching to reduce confounding factors. The resulting drug-AE associations are compared to those from the Side Effect Resource, which are primarily derived from clinical trials. The analysis reveals that the datasets generally share less than 10% of reported AEs for the same drug and have different distributions of AEs across System Organ Classes (SOCs). Using the drugs from the two AE datasets described in the first chapter, the second chapter integrates corresponding bioactivities, i.e. measured potencies and affinities from the ChEMBL database and ligand-based target predictions obtained with the tool PIDGIN, with drug plasma concentrations compiled from literature, such as Cmax. Compared to a constant bioactivity cut-off of 1 uM, using the ratio of the unbound drug plasma concentration over the drug potency, i.e. Cmax/XC50, results in different binary activity calls for protein targets. Whether deriving activity calls in this way results in the selection of targets with greater relevance to human AEs is investigated in the third chapter, which computes relationships between targets and AEs using different measures of statistical association. Using the Cmax/XC50 ratio results in higher Likelihood Ratios and Positive Predictive Values (PPVs) for target-AE associations that were previously reported in the context of secondary pharmacology screening, at the cost of a lower recall, possibly due to the smaller size of the dataset with available plasma concentrations. Furthermore, a large-scale quantitative assessment of bioactivities as indicators of AEs reveals a trade-off between the PPV and how many AE-associated drugs can potentially be detected from in vitro screening, although using combinations of targets can improve the detection rate in ~40% of cases at limited cost to the PPV. The work highlights AEs most strongly related to bioactivities and their SOC distribution. Overall, this thesis contributes to knowledge of the relationships between in vitro bioactivities and empirical evidence of AEs in humans. The results can inform the selection of proteins for secondary pharmacology screening and the development of computational models to predict AEs.