Post-translational modifications by the Small Ubiquitin-like Modifier (SUMO) are essential for many eukaryotic cellular functions. Several large-scale experimental datasets and sequence-based predictions exist that identify SUMOylated proteins. However, the overlap between these datasets is small, suggesting many false positives with low functional relevance. Therefore, we applied machine learning techniques to a diverse set of large-scale SUMOylation studies combined with protein characteristics such as cellular function and protein-protein interactions, to provide integrated SUMO predictions for human and yeast cells (iSUMO). Protein-protein and protein-nucleic acid interactions prove to be highly predictive of protein SUMOylation, supporting a role of the modification in protein complex formation. We note the marked prevalence of SUMOylation amongst RNA-binding proteins. We predict 1,596 and 492 SUMO targets in human and yeast, respectively (5% false positive rate, FPR), which is five times more than what existing sequence-based tools predict at the same FPR. One third of the predictions are validated by an independent, high-quality dataset. iSUMO therefore represents a comprehensive SUMO prediction tool for human and yeast with a high probability for functional relevance of the predictions.