Ubiquitin-specific-processing proteases (USPs), the largest deubiquitinating enzyme (DUB) subfamily, play critical roles in cancer. However, clinical utility of USPs is hindered by limited knowledge about their varied and substrate-dependent actions. Here, we performed a comprehensive investigation on pan-cancer impacts of USPs by integrating multi-omics data and annotated data resources, especially a deubiquitination network. Meaningful insights into the roles of 54 USPs in 29 types of cancers were generated. Although rare mutations were observed, a majority of USPs exhibited significant expressional alterations, prognostic impacts and strong correlations with cancer hallmark pathways. Notably, from our DUB-substrate interaction prediction model, additional USP-substrate interactions (USIs) were recognized to complement knowledge gap about cancer-relevant USIs. Intriguingly, expression signatures of the USIs revealed clinically meaningful cancer subtypes, where key USPs and substrates cooperatively contributed to significant prognosis differences among subtypes. Overall, this investigation provides a valuable resource to assist mechanism research and clinical utility about USPs.